1#[derive(Clone, Copy, Debug, PartialEq, Eq)]
2pub enum Oracle {
3 Hercules,
5 EnterpriseCobol,
7 Db2,
9}
10
11#[derive(Clone, Copy, Debug, PartialEq, Eq)]
12pub enum Basis {
13 Documented,
15 Recalled,
17 Chosen,
19 Observed,
22}
23
24#[derive(Clone, Copy, Debug)]
25pub struct Assumption {
26 pub id: &'static str,
27 pub claim: &'static str,
28 pub basis: Basis,
29 pub oracle: Oracle,
30}
31
32pub const HFP_EXTENDED_LOW_HALF: &str = "M1";
33pub const HFP_FROM_FIXED_TRUNCATES: &str = "M2";
34pub const INTERMEDIATE_TABLE: &str = "C1";
35pub const TRUNC_OPT_IS_BINARY: &str = "C2";
36pub const PFD_MOVES_BYTES: &str = "C3";
37pub const PFD_COMPARES_LOGICALLY: &str = "C4";
38pub const FLOAT_FROM_DECIMAL: &str = "C5";
39pub const FLOAT_TO_DECIMAL: &str = "C6";
40pub const FLOAT_NARROWING_ROUNDS: &str = "C7";
41pub const LE_MASKS_UNDERFLOW: &str = "C8";
42pub const PREFERRED_RESULT_SIGNS: &str = "C9";
43pub const ZONED_BY_PACK: &str = "C10";
44pub const CCSID_TABLES: &str = "C11";
45pub const WORKING_STORAGE_LAYOUT: &str = "C12";
46pub const NOPFD_REPAIRS_UNSIGNED_INPUT: &str = "C13";
47pub const DISPLAY_OF_NONDISPLAY_NUMERIC: &str = "C14";
48pub const ACCEPT_AT_END: &str = "C15";
49pub const SYSIN_CARD_IMAGES: &str = "C261";
50pub const CONTENT_LITERAL_ZONED: &str = "C16";
51pub const KEYED_FILE_STATUS: &str = "C17";
52pub const ALTERNATE_KEY_ORDER: &str = "C18";
53pub const FAILED_READ_LOSES_POSITION: &str = "C19";
54pub const START_COMPARES_SHORTER: &str = "C20";
55pub const REWRITE_SHARED_ALTERNATE: &str = "C21";
56pub const CICS_FRESH_STORAGE: &str = "C22";
57pub const CICS_CALEN_RESTORED: &str = "C23";
58pub const CICS_HANDLE_ABEND_CATCHES_CONDITIONS: &str = "C24";
59pub const CICS_LENGTH_DEFAULTS_TO_INTO: &str = "C25";
60pub const CICS_PROGRAM_CHECK_IS_ASRA: &str = "C26";
61pub const CICS_BROWSE_SKIP: &str = "C27";
62pub const BMS_RECEIVE_NULLS: &str = "C28";
63pub const BMS_INPUT_JUSTIFY: &str = "C29";
64pub const BMS_EXTENDED_ORDER: &str = "C30";
65pub const BMS_CONSTANTS_UNVERIFIED: &str = "C31";
66pub const BMS_SEND_DATA_CHOICE: &str = "C32";
67pub const CICS_INITIAL_AID: &str = "C33";
68pub const LONG_ZONED_BY_PACKS: &str = "C34";
69pub const ASCII_COLLATION: &str = "C35";
70pub const TABLE_SORT_COLLATION: &str = "C36";
71pub const ALPHABET_LITERALS: &str = "C37";
72pub const LE_SERVICE_AFTER_PROGRAMS: &str = "L1";
73pub const LE_ARGUMENTS_BY_ADDRESS: &str = "L2";
74pub const LE_FEEDBACK_TOKEN: &str = "L3";
75pub const LE_FEEDBACK_NO_INSTANCE_INFO: &str = "L4";
76pub const LE_OMITTED_FC_ABENDS: &str = "L5";
77pub const LE_CEE3ABD: &str = "L6";
78pub const LE_LOCAL_TIME_IS_UTC: &str = "L7";
79pub const LE_SECONDS_HFP: &str = "L8";
80pub const LE_PICTURE_OUTPUT: &str = "L9";
81pub const LE_PICTURE_INPUT: &str = "L10";
82pub const LE_CENTURY_WINDOW: &str = "L11";
83pub const LE_RETURN_CODE_UNCHANGED: &str = "L12";
84pub const LE_MESSAGE_AND_DUMP_FILES: &str = "L13";
85pub const LE_HEAP: &str = "L14";
86pub const LE_UNDER_CICS: &str = "L15";
87pub const LE_CEEIGZCT: &str = "L16";
88pub const LE_CEEIGZCT_DISAGREEMENTS: &str = "L17";
89pub const LE_SHORT_ARGUMENT_LIST: &str = "L18";
90pub const LE_SSRANGE_U4038: &str = "L19";
91pub const REPORT_WRITER_PRECOMPILER: &str = "RW1";
92pub const REPORT_TOTALS_BEFORE_PAGE_FIT: &str = "RW2";
93pub const REPORT_SOURCE_SUM_CORRELATION: &str = "RW3";
94pub const REPORT_PAGE_REGION_DEFAULTS: &str = "RW4";
95pub const REPORT_LINE_WRITES: &str = "RW5";
96pub const REPORT_CARRIAGE_CONTROL: &str = "RW6";
97pub const REPORT_RECORD_LENGTH: &str = "RW7";
98pub const REPORT_SUM_OVERFLOW: &str = "RW8";
99pub const REPORT_SOURCE_OVERFLOW: &str = "RW9";
100pub const REPORT_SUPPRESS_PRINTING: &str = "RW10";
101pub const REPORT_NEW_PAGES: &str = "RW11";
102pub const REPORT_OUT_OF_ORDER: &str = "RW12";
103pub const REPORT_CONTROL_AREA: &str = "RW13";
104pub const OBJECT_REFERENCE_VALUE: &str = "J1";
105pub const LOCAL_REFERENCES_EXPIRE: &str = "J2";
106pub const OBJECTS_NEVER_FREED: &str = "J3";
107pub const INSTANCE_DATA_START: &str = "J4";
108pub const FACTORY_DATA_START: &str = "J5";
109pub const METHOD_LOOKUP: &str = "J6";
110pub const METHOD_NAME_ITEM: &str = "J7";
111pub const INVOKE_NULL: &str = "J8";
112pub const NO_METHOD_ABEND: &str = "J9";
113pub const FACTORY_SELF: &str = "J10";
114pub const CLASS_SEARCH: &str = "J11";
115pub const INVOKE_KEEPS_RETURN_CODE: &str = "J12";
116pub const OO_OPTIONS_REQUIRED: &str = "J13";
117pub const CHAR_FROM_DISPLAY: &str = "J14";
118pub const LOCAL_REFERENCES_OUTSIDE_METHODS: &str = "J15";
119pub const SELF_IS_LOCAL: &str = "J16";
120pub const EXPIRED_REFERENCE_ABENDS: &str = "J17";
121pub const LOCAL_FRAMES: &str = "J18";
122pub const OO_OPTIONS_SEVERITY: &str = "J19";
123pub const REFERENCES_KEPT: &str = "J20";
124pub const SQL_COMMIT_AT_NORMAL_END: &str = "SQ1";
125pub const SQL_WHENEVER_ORDER: &str = "SQ2";
126pub const SQL_TRUNCATED_INDICATOR: &str = "SQ3";
127pub const SQL_POSTGRES_ERRORS: &str = "SQ4";
128pub const SQL_DIALECT_REWRITES: &str = "SQ5";
129pub const SQL_INTO_WITHOUT_COLONS: &str = "SQ7";
130pub const SQL_DECLARATIONS_CROSS_NESTED_PROGRAMS: &str = "SQ8";
131pub const SQL_DOUBLE_TO_HFP_TRUNCATES: &str = "SQ9";
132pub const SQL_ZONED_IS_DECIMAL: &str = "SQ10";
133pub const SQL_ISO_DATETIME: &str = "SQ11";
134pub const SQL_TRAILING_BLANKS_SENT: &str = "SQ12";
135pub const SQL_FETCH_ROW_COUNT: &str = "SQ13";
136pub const SORT_EQUAL_KEYS_IN_ORDER: &str = "S1";
137pub const MERGE_EQUAL_KEYS_BY_FILE: &str = "S2";
138pub const MERGE_OUT_OF_SEQUENCE_FAILS: &str = "S3";
139pub const SORT_FILE_FAILURE: &str = "S4";
140pub const SORT_DECIMAL_KEYS: &str = "S5";
141pub const SORT_RECORD_LENGTHS: &str = "S6";
142pub const SORT_RETURN_STOPS: &str = "S7";
143pub const RETURN_AFTER_END: &str = "S8";
144pub const SORT_KEY_INVALID_DIGITS: &str = "S9";
145pub const SORT_NEGATIVE_ZERO: &str = "S10";
146pub const FASTSRT_FILES: &str = "S11";
147pub const FASTSRT_STATUS: &str = "S12";
148pub const FASTSRT_FAILURE: &str = "S13";
149pub const SAME_AREA_VSAM: &str = "S14";
150pub const FASTSRT_PRINT_RECORDS: &str = "S15";
151pub const FASTSRT_RECORD_LENGTHS: &str = "S16";
152pub const FASTSRT_ADV_PRINT: &str = "S17";
153pub const PRINT_CONTROL_CHARACTER: &str = "C40";
154pub const PRINT_SPACING_RECORDS: &str = "C41";
155pub const PRINT_CONTROL_RUN_TIME: &str = "C42";
156pub const TEXT_PRINT_LINES: &str = "C43";
157pub const COMPILER_SEVERITIES: &str = "C44";
158pub const REFUSALS_ARE_SEVERE: &str = "C45";
159pub const REFUSED_FROM_S: &str = "C46";
160pub const NOCOMPILE: &str = "C47";
161pub const COMMENT_ENTRY_EXTENT: &str = "C80";
162pub const COMMENT_ENTRY_HEADERS: &str = "C81";
163pub const COMMENT_ENTRY_REMARKS: &str = "C82";
164pub const CONTINUED_LITERAL_QUOTES: &str = "C83";
165pub const COPY_SEARCH_ROUNDS: &str = "C85";
166pub const COPY_NOT_THE_PROGRAM: &str = "C86";
167pub const COPY_LITERAL_AS_WRITTEN: &str = "C87";
168pub const COPY_DOUBLED_PERIOD: &str = "C88";
169pub const LINAGE_COUNTER: &str = "C70";
170pub const LINAGE_PAGE_MOVEMENT: &str = "C71";
171pub const LINAGE_END_OF_PAGE: &str = "C72";
172pub const LINAGE_VALUES: &str = "C73";
173pub const LINAGE_EXTEND: &str = "C74";
174pub const LINAGE_COUNTER_BETWEEN_WRITES: &str = "C75";
175pub const PRINT_FILE_UPDATE: &str = "C76";
176pub const ENTRY_IN_SEQUENCE: &str = "C50";
177pub const ENTRY_CALLS: &str = "C51";
178pub const ALTERED_GO_TO_RESET: &str = "C52";
179pub const DISPLAY_STREAM: &str = "C53";
180pub const RANDOM_GENERATOR: &str = "C54";
181pub const ZERO_DIVISOR_CHECK: &str = "C55";
182pub const ERROR_DECLARATIVE_MODE: &str = "C60";
183pub const ERROR_DECLARATIVE_STATUSES: &str = "C61";
184pub const SORT_FILE_DECLARATIVE: &str = "C62";
185pub const DEBUG_RUNTIME_OPTION: &str = "C63";
186pub const DEBUG_LINE_NUMBER: &str = "C64";
187pub const DEBUG_LINE_STATEMENT: &str = "C65";
188pub const DEBUG_CONTENTS_LENGTH: &str = "C66";
189pub const DEBUG_NAME_FORM: &str = "C67";
190pub const DEBUGGING_SECTION_REFERENCES: &str = "C68";
191pub const GLOBAL_DECLARATIVES: &str = "C69";
192pub const SYNC_SUBORDINATE_GROUP: &str = "C90";
193pub const SYNC_ONLY_WHEN_WRITTEN: &str = "C91";
194pub const SYNC_SLACK_OWNER: &str = "C92";
195pub const SYNC_REDEFINES_REFUSED: &str = "C93";
196pub const DECIMAL_COMMA_SEPARATOR: &str = "C94";
197pub const DECIMAL_COMMA_DISPLAY_LITERAL: &str = "C95";
198pub const ARITH_DIGIT_LIMITS: &str = "C96";
199pub const MULTIPLE_RESULTS: &str = "C97";
200pub const ALTER_DEBUGGING: &str = "C98";
201pub const PERFORM_RETURN_POINTS: &str = "C99";
202pub const FLOAT_FUNCTION_ARGUMENTS: &str = "C100";
203pub const FLOAT_FUNCTION_ROUNDING: &str = "C110";
204pub const FLOATING_POINT_FUNCTIONS: &str = "C111";
205pub const FUNCTION_DOMAIN: &str = "C112";
206pub const NUMVAL_TEST_RULES: &str = "C113";
207pub const FUNCTION_CLOCK: &str = "C114";
208pub const UUID4_SOURCE: &str = "C115";
209pub const FORMATTED_DATETIME_RULES: &str = "C116";
210pub const ROUNDED_EXTRA_PLACE: &str = "C101";
211pub const CURRENCY_SIGNS: &str = "C102";
212pub const JSON_GENERATE_RULES: &str = "C117";
213pub const XML_PARSE_RULES: &str = "C118";
214pub const XML_GENERATE_RULES: &str = "C119";
215pub const JSON_PARSE_RULES: &str = "C200";
216pub const DLI_TRANSLATION: &str = "C201";
217pub const PASSWORD_IGNORED: &str = "C202";
218pub const CORRESPONDING_PAIRS: &str = "C130";
219pub const CORRESPONDING_CHOICES: &str = "C131";
220pub const STOP_LITERAL: &str = "C132";
221pub const NUMPROC_MIG_WARNS: &str = "C120";
222pub const INVALID_OPTION_DISCARDED: &str = "C121";
223pub const OPTIONS_WITHOUT_EFFECT: &str = "C122";
224pub const NON_COBOL_CHARACTERS: &str = "C123";
225pub const NO_PROGRAM_END: &str = "C124";
226pub const USE_WITHOUT_PARAGRAPH: &str = "C125";
227pub const PICTURE_ENDS_AT_ITS_SEPARATOR: &str = "C195";
228pub const ZONED_COMPARED_AS_BYTES: &str = "C221";
229pub const APOST_EVERYWHERE: &str = "C210";
230pub const CURRENCY_OPTION: &str = "C211";
231pub const NSYMBOL_DBCS: &str = "C212";
232pub const INITIAL_UNDER_THREAD: &str = "C217";
233pub const VLR_WITHOUT_VARYING: &str = "C218";
234pub const VLR_RECORDS_CHECKED: &str = "C219";
235pub const VSAM_DATA_SET_LEFT_OPEN: &str = "C220";
236pub const DISPSIGN_SEPARATE: &str = "C213";
237pub const LILIAN_INTEGER_DATES: &str = "C214";
238pub const CEECBLDY_UNDER_LILIAN: &str = "C215";
239pub const COMPLETE_SET_OF_QUALIFIERS: &str = "C216";
240pub const INSPECT_FUNCTION_SUBJECT: &str = "C190";
241pub const INSPECT_NATIONAL_FUNCTION_RESULT: &str = "C191";
242pub const NATIONAL_CASE_AND_REVERSE: &str = "C192";
243pub const INVDATA_CLEANSIGN: &str = "C222";
244pub const INVDATA_ZONES_COMPARED: &str = "C223";
245pub const OPTIMIZED_ZONES_COMPARED: &str = "C262";
246pub const ALPHANUMERIC_MOVED_UNCHECKED: &str = "C240";
247pub const NUMERIC_MOVED_UNCHECKED: &str = "C260";
248pub const STATEMENT_LIMIT_IS_TIME: &str = "C241";
249pub const INITCHECK_ANALYSIS: &str = "C224";
250pub const INITCHECK_MESSAGE: &str = "C225";
251pub const NUMCHECK_SENDERS: &str = "C228";
252pub const NUMCHECK_MESSAGE: &str = "C229";
253pub const NUMCHECK_LAX_REDEFINES: &str = "C280";
254pub const NUMCHECK_ALWAYS_FAILS: &str = "C281";
255pub const DBCS_UNDER_SINGLE_BYTE_PAGE: &str = "C282";
256pub const DBCS_LITERAL_SOURCE: &str = "C283";
257pub const MIXED_PAGE_DATA: &str = "C284";
258pub const DBCS_DISPLAY: &str = "C285";
259pub const DBCS_HOST_VARIABLES: &str = "C286";
260pub const PARMCHECK_BUFFER: &str = "C226";
261pub const PARMCHECK_MESSAGE: &str = "C227";
262pub const PARM_ARGUMENTS_BEFORE_LAST_SLASH: &str = "C250";
263pub const PARM_AREA_PADDED: &str = "C251";
264pub const ABBREVIATED_RELATIONS: &str = "C150";
265pub const PARAGRAPH_IN_OWN_SECTION: &str = "C151";
266pub const REPLACE_STATEMENT: &str = "C160";
267pub const VARIABLY_LOCATED_ITEMS: &str = "C161";
268pub const EXTERNAL_STORAGE: &str = "C180";
269pub const GLOBAL_NAMES: &str = "C181";
270pub const SET_TO_ENTRY: &str = "C140";
271pub const HEX_CURRENCY_SIGN: &str = "C141";
272pub const INITIALIZE_FLOAT_NUMERIC: &str = "C171";
273pub const FLOAT_VALUE_LITERAL: &str = "C172";
274pub const CICS_ABEND_EXITS: &str = "C142";
275pub const FUNCTION_SOURCE_ORDER: &str = "C270";
276pub const FUNCTION_NAMED_AS_INTRINSIC: &str = "C271";
277pub const FUNCTION_ARGUMENT_TEMPORARIES: &str = "C272";
278pub const FUNCTION_SQL_CICS: &str = "C273";
279pub const FUNCTION_INVOCATION_ORDER: &str = "C274";
280pub const INSPECT_NATIONAL_ITEM: &str = "C230";
281pub const INSPECT_OPERAND_USAGE: &str = "C231";
282pub const INSPECT_OPERAND_MADE_NATIONAL: &str = "C232";
283pub const CICS_ABEND_LABEL_GO_TO: &str = "C236";
284pub const CICS_ABEND_EXIT_ACROSS_CALL: &str = "C237";
285pub const CICS_ABEND_LABEL_OWNER: &str = "C238";
286pub const CICS_ABEND_EXIT_ACROSS_XCTL: &str = "C239";
287pub const CICS_RETURN_ENDS_THE_LEVEL: &str = "C233";
288pub const CICS_HANDLERS_ACROSS_CALL: &str = "C234";
289pub const CICS_CONDITION_LABEL_OWNER: &str = "C235";
290pub const CICS_RETURN_BELOW_THE_FIRST_LEVEL: &str = "C143";
291pub const CICS_STOP_RUN_ENDS_THE_LEVEL: &str = "C144";
292pub const CICS_RUN_UNIT_PER_LINK: &str = "C145";
293pub const CICS_HANDLERS_ACROSS_XCTL: &str = "C146";
294pub const CICS_NO_OBJECT_ORIENTED_COBOL: &str = "C147";
295pub const CICS_TRANSFER_TO_A_RUNNING_PROGRAM: &str = "C148";
296pub const CICS_ENCLAVE_EXTERNALS_AND_HEAP: &str = "C126";
297pub const RECURSIVE_CALL_OF_AN_ACTIVE_PROGRAM: &str = "C127";
298pub const CICS_RETURN_COMMAREA_LENGTH: &str = "C128";
299pub const CICS_RUN_UNIT_STORAGE_RELEASED: &str = "C129";
300pub const TRAP_OFF_LEAVES_FILES_OPEN: &str = "C152";
301pub const CICS_TRANSFER_COMMAREA_LENGTH: &str = "C103";
302pub const CICS_RANDOM_PER_RUN_UNIT: &str = "C104";
303pub const CICS_RETURN_CODE_PER_RUN_UNIT: &str = "C105";
304pub const CICS_ENTRY_POINTERS_ACROSS_RUN_UNITS: &str = "C106";
305pub const INITIALIZE_REFERENCE_MODIFIED: &str = "C300";
306pub const SORT_INVALID_DIGIT_ABENDS: &str = "C340";
307pub const SORT_IFTHEN_FIXED_LENGTH: &str = "C341";
308pub const SORT_MASK_GROUPS_OF_THREE: &str = "C342";
309pub const SORT_PATTERN_DECIMAL_POINT: &str = "C343";
310pub const ALTERNATE_KEYS_FROM_THE_BASE: &str = "C350";
311pub const PATH_PRESENTS_THE_BASE: &str = "C351";
312pub const GENERATED_COMPONENT_NAMES: &str = "C352";
313pub const BLDINDEX_NON_ENDING_ERRORS: &str = "C353";
314pub const LISTCAT_WHAT_THE_CATALOG_KEEPS: &str = "C354";
315pub const UPGRADE_SET_AFTER_THE_STEP: &str = "C355";
316pub const BLDINDEX_REFUSALS: &str = "C356";
317pub const PRINT_LISTING_LAYOUT: &str = "C358";
318pub const PRINT_RANGE_ENDS: &str = "C359";
319pub const INSPECT_SIGNED_ZONED: &str = "C330";
320pub const RELATIVE_NUMBER_BELOW_ONE: &str = "C331";
321pub const DISPLAY_NUMERIC_FUNCTION: &str = "C332";
322pub const BY_VALUE_TO_REFERENCE: &str = "C333";
323pub const FLOAT_EXPONENTIATION: &str = "C334";
324
325pub const ASSUMPTIONS: &[Assumption] = &[
326 Assumption {
327 id: HFP_EXTENDED_LOW_HALF,
328 claim: "An extended HFP result's low half carries the high characteristic minus 14, modulo 128, and is all zero when the value is all zero bits (SA22-7832-14, HFP extended format, p. 18-4)",
329 basis: Basis::Documented,
330 oracle: Oracle::Hercules,
331 },
332 Assumption {
333 id: HFP_FROM_FIXED_TRUNCATES,
334 claim: "CONVERT FROM FIXED to HFP truncates the hexadecimal digits the precision cannot hold",
335 basis: Basis::Recalled,
336 oracle: Oracle::Hercules,
337 },
338 Assumption {
339 id: INTERMEDIATE_TABLE,
340 claim: "An operation's intermediate result has i integer and d decimal places: for + and -, one integer place more than the operand with more, and the more decimal places; for *, the sum of each; for /, the dividend's integer places and the divisor's decimal places together, and the dividend's decimal places less the divisor's or dmax, whichever is more. Up to 30 digits (31 under ARITH(EXTEND)) i and d are carried; beyond that, N-d and d when d <= dmax, else i and N-i when i+dmax <= N, else N-dmax and dmax; digits beyond are truncated (Programming Guide SC27-8714-03, pp. 794-795)",
341 basis: Basis::Documented,
342 oracle: Oracle::EnterpriseCobol,
343 },
344 Assumption {
345 id: TRUNC_OPT_IS_BINARY,
346 claim: "Under TRUNC(OPT) a binary receiver whose value exceeds its PICTURE is truncated at the halfword, fullword or doubleword, as under TRUNC(BIN)",
347 basis: Basis::Chosen,
348 oracle: Oracle::EnterpriseCobol,
349 },
350 Assumption {
351 id: PFD_MOVES_BYTES,
352 claim: "Under NUMPROC(PFD) a MOVE between packed items of the same length and scale copies the bytes, so a non-preferred sign passes through; under NOPFD the receiver gets the preferred sign",
353 basis: Basis::Chosen,
354 oracle: Oracle::EnterpriseCobol,
355 },
356 Assumption {
357 id: PFD_COMPARES_LOGICALLY,
358 claim: "Under NUMPROC(PFD) two packed items of the same length and scale compare byte by byte, so X'1F' and X'1C' differ",
359 basis: Basis::Chosen,
360 oracle: Oracle::EnterpriseCobol,
361 },
362 Assumption {
363 id: FLOAT_FROM_DECIMAL,
364 claim: "A fixed-point decimal m with s decimal places becomes HFP as CONVERT FROM FIXED of m, then an HFP divide by 10^s",
365 basis: Basis::Chosen,
366 oracle: Oracle::EnterpriseCobol,
367 },
368 Assumption {
369 id: FLOAT_TO_DECIMAL,
370 claim: "An HFP value moved or stored into a fixed-point receiver is rounded in the receiver's low-order position, with or without ROUNDED, and keeps at most 9 significant digits from short precision and 18 from long, the rest zero (Programming Guide SC27-8714-03, p. 52); the Language Reference's COMBINED-DATETIME example, 143951.1886781248 from a long value of 143951.18867812478..., shows a COMPUTE rounding (SC27-8713-03, p. 542). Rounding is half away from zero; an extended value keeps every digit the receiver holds",
371 basis: Basis::Documented,
372 oracle: Oracle::EnterpriseCobol,
373 },
374 Assumption {
375 id: FLOAT_NARROWING_ROUNDS,
376 claim: "A floating-point value moved or stored into a narrower COMP-1 or COMP-2 is rounded in the low-order position, as LOAD ROUNDED rounds (Programming Guide SC27-8714-03, p. 52)",
377 basis: Basis::Documented,
378 oracle: Oracle::EnterpriseCobol,
379 },
380 Assumption {
381 id: LE_MASKS_UNDERFLOW,
382 claim: "Language Environment runs COBOL with the HFP exponent-underflow and significance masks off, so both yield a true zero",
383 basis: Basis::Chosen,
384 oracle: Oracle::EnterpriseCobol,
385 },
386 Assumption {
387 id: PREFERRED_RESULT_SIGNS,
388 claim: "Arithmetic results take the preferred sign under either NUMPROC setting: C or D for a signed item, F for an unsigned one",
389 basis: Basis::Recalled,
390 oracle: Oracle::EnterpriseCobol,
391 },
392 Assumption {
393 id: ZONED_BY_PACK,
394 claim: "A zoned operand enters arithmetic as PACK leaves it: zones other than the sign's are discarded, so an embedded space is a zero digit",
395 basis: Basis::Chosen,
396 oracle: Oracle::EnterpriseCobol,
397 },
398 Assumption {
399 id: CCSID_TABLES,
400 claim: "The compiler's conversion to UTF-16 for each carried CCSID is ICU's ibm-* table for it, byte for byte",
401 basis: Basis::Chosen,
402 oracle: Oracle::EnterpriseCobol,
403 },
404 Assumption {
405 id: WORKING_STORAGE_LAYOUT,
406 claim: "Each 01 and 77 item of WORKING-STORAGE starts on an 8-byte boundary in source order, and storage without a VALUE starts as X'00'",
407 basis: Basis::Chosen,
408 oracle: Oracle::EnterpriseCobol,
409 },
410 Assumption {
411 id: NOPFD_REPAIRS_UNSIGNED_INPUT,
412 claim: "Under NUMPROC(NOPFD) an unsigned packed or zoned operand's sign is forced to X'F' before use, so no sign makes it a data exception; under PFD it is used as it stands",
413 basis: Basis::Chosen,
414 oracle: Oracle::EnterpriseCobol,
415 },
416 Assumption {
417 id: DISPLAY_OF_NONDISPLAY_NUMERIC,
418 claim: "DISPLAY shows a packed or binary item as zoned digits of its PICTURE, a negative value's sign overpunched on the last digit and a positive value's digits unsigned (Programming Guide SC27-8714-03, p. 363, Table 48; Language Reference SC27-8713-03, p. 334); a COMP-5 item, or any binary item under TRUNC(BIN), shows its whole binary value in 5, 10, or 19 (signed) or 20 digits for a halfword, fullword or doubleword. Under --dialect gnucobol DISPLAY shows them as cobc -std=ibm does, whatever DISPSIGN says: a signed item's sign, + or -, before its digits, a packed item's digits those of its PICTURE, and any binary item's whole value in 5, 10 or 20 digits for a halfword, fullword or doubleword",
419 basis: Basis::Chosen,
420 oracle: Oracle::EnterpriseCobol,
421 },
422 Assumption {
423 id: ACCEPT_AT_END,
424 claim: "ACCEPT from SYSIN at its end leaves the receiving item unchanged and the run continues. Under --dialect gnucobol the item takes a space, as cobc moves one at the end of its input, so a numeric or numeric-edited item becomes zero and any other is filled with spaces",
425 basis: Basis::Chosen,
426 oracle: Oracle::EnterpriseCobol,
427 },
428 Assumption {
429 id: SYSIN_CARD_IMAGES,
430 claim: "ACCEPT from SYSIN reads each line of the SYSIN file as one record, 80 bytes as a card of in-stream data is: a shorter line is padded with spaces to 80, and a longer line is a record of its own length, as a variable-length record would be. The receiving item is filled from consecutive records with no conversion, editing or check, the last record cut where the item ends; at the end of SYSIN after some data the rest of the item is spaces, and before any data the item is unchanged (C15) (Language Reference SC27-8713-03, pp. 307-308: 'There is no editing or error checking of the incoming data', each record concatenated with the previous, a fixed-length record used whole). A numeric receiver takes the characters as they are, so an empty line gives spaces and a non-digit is a data exception only where the item is next read as a number. The record length of a SYSIN data set other than in-stream cards is not known to ironwork, and 80 is chosen. A national receiver still takes one line, converted from the code page, where IBM takes UTF-16 data unconverted",
431 basis: Basis::Chosen,
432 oracle: Oracle::EnterpriseCobol,
433 },
434 Assumption {
435 id: CONTENT_LITERAL_ZONED,
436 claim: "A numeric literal passed BY CONTENT arrives as zoned decimal of its own digits, the sign overpunched when negative",
437 basis: Basis::Chosen,
438 oracle: Oracle::EnterpriseCobol,
439 },
440 Assumption {
441 id: KEYED_FILE_STATUS,
442 claim: "Indexed and relative files report the file status values of Table 34 (Language Reference SC27-8713-03, pp. 300-302): 02 only for an indexed file with an alternate key that allows duplicates, when a READ finds the next record by the key of reference has the same key or a WRITE or REWRITE makes a duplicate alternate key value, so never for a relative file; 14 for a sequential READ of a relative file whose record number has more digits than the RELATIVE KEY; 21 for a sequentially accessed indexed file's WRITE whose prime key is not above the last, or a REWRITE that changed the prime key; 22 for a duplicate prime key, relative record number, or alternate key without DUPLICATES; 23 for no such record; 24 for a WRITE beyond the file's boundaries, or a sequential WRITE whose relative record number has more digits than the RELATIVE KEY; 43 for a sequential-access REWRITE or DELETE whose last input-output statement was not a successful READ; 46 for a sequential READ with no valid next record; and 47, 48 and 49 for the wrong open mode",
443 basis: Basis::Documented,
444 oracle: Oracle::EnterpriseCobol,
445 },
446 Assumption {
447 id: ALTERNATE_KEY_ORDER,
448 claim: "Records sharing an alternate key come back in the order they were written; a record REWRITTEN with a new alternate key goes after the others with it",
449 basis: Basis::Recalled,
450 oracle: Oracle::EnterpriseCobol,
451 },
452 Assumption {
453 id: FAILED_READ_LOSES_POSITION,
454 claim: "A random READ that finds no record leaves no next record, so READ NEXT then gives 46",
455 basis: Basis::Chosen,
456 oracle: Oracle::EnterpriseCobol,
457 },
458 Assumption {
459 id: START_COMPARES_SHORTER,
460 claim: "START compares the key with its operand over the shorter of the two, left to right, as bytes",
461 basis: Basis::Recalled,
462 oracle: Oracle::EnterpriseCobol,
463 },
464 Assumption {
465 id: CICS_FRESH_STORAGE,
466 claim: "A COBOL program reached by EXEC CICS LINK or XCTL starts with fresh WORKING-STORAGE each time, unlike one reached by CALL",
467 basis: Basis::Recalled,
468 oracle: Oracle::EnterpriseCobol,
469 },
470 Assumption {
471 id: CICS_CALEN_RESTORED,
472 claim: "EIBCALEN is the linked program's COMMAREA length while it runs and the caller's again after the LINK returns",
473 basis: Basis::Chosen,
474 oracle: Oracle::EnterpriseCobol,
475 },
476 Assumption {
477 id: CICS_HANDLE_ABEND_CATCHES_CONDITIONS,
478 claim: "A HANDLE ABEND exit receives control when a condition nothing handles abends the task (AEIx), and when a condition whose HANDLE CONDITION label another program set abends it APC2 (C235), as for any abend it can intercept, and is deactivated by being taken (C142)",
479 basis: Basis::Documented,
480 oracle: Oracle::EnterpriseCobol,
481 },
482 Assumption {
483 id: CICS_LENGTH_DEFAULTS_TO_INTO,
484 claim: "READ, READNEXT, READPREV, READQ TS and READQ TD without LENGTH take the INTO item's length as the limit, so a longer record raises LENGERR",
485 basis: Basis::Recalled,
486 oracle: Oracle::EnterpriseCobol,
487 },
488 Assumption {
489 id: CICS_PROGRAM_CHECK_IS_ASRA,
490 claim: "A program check (S0C4, S0C7 and the like) in a CICS task ends it with transaction abend ASRA",
491 basis: Basis::Recalled,
492 oracle: Oracle::EnterpriseCobol,
493 },
494 Assumption {
495 id: CICS_BROWSE_SKIP,
496 claim: "READNEXT after the program changed RIDFLD to a key the browse is not at continues from the first record at or after the new RIDFLD (skip-sequential)",
497 basis: Basis::Recalled,
498 oracle: Oracle::EnterpriseCobol,
499 },
500 Assumption {
501 id: BMS_RECEIVE_NULLS,
502 claim: "RECEIVE MAP sets the input map to nulls, then fills only the fields the operator modified; a field erased to empty gets F = X'80' and L = 0",
503 basis: Basis::Recalled,
504 oracle: Oracle::EnterpriseCobol,
505 },
506 Assumption {
507 id: BMS_INPUT_JUSTIFY,
508 claim: "Input data lands left-justified and blank-filled unless JUSTIFY says otherwise, and a NUM field right-justified and zero-filled, the defaults IBM documents for JUSTIFY",
509 basis: Basis::Recalled,
510 oracle: Oracle::EnterpriseCobol,
511 },
512 Assumption {
513 id: BMS_EXTENDED_ORDER,
514 claim: "A field's extended attribute bytes in the symbolic map follow its A byte in the order COLOR, PS, HILIGHT, VALIDN, OUTLINE, SOSI, TRANSP",
515 basis: Basis::Chosen,
516 oracle: Oracle::EnterpriseCobol,
517 },
518 Assumption {
519 id: BMS_CONSTANTS_UNVERIFIED,
520 claim: "DFHNULL is X'00', and DFHBMPEM, DFHBMPNL, DFHBMPFF and DFHBMPCR are X'19', X'15', X'0C' and X'0D'",
521 basis: Basis::Recalled,
522 oracle: Oracle::EnterpriseCobol,
523 },
524 Assumption {
525 id: BMS_SEND_DATA_CHOICE,
526 claim: "SEND MAP without MAPONLY or DATAONLY sends a field's symbolic data when its first byte is not X'00', else the map's INITIAL, and a non-null A byte replaces ATTRB",
527 basis: Basis::Recalled,
528 oracle: Oracle::EnterpriseCobol,
529 },
530 Assumption {
531 id: CICS_INITIAL_AID,
532 claim: "A task started by terminal input sees that input's AID in EIBAID before any RECEIVE",
533 basis: Basis::Recalled,
534 oracle: Oracle::EnterpriseCobol,
535 },
536 Assumption {
537 id: REWRITE_SHARED_ALTERNATE,
538 claim: "REWRITE gives 02 whenever another record shares one of the record's alternate keys that allow duplicates, whether or not that key changed",
539 basis: Basis::Chosen,
540 oracle: Oracle::EnterpriseCobol,
541 },
542 Assumption {
543 id: LE_SERVICE_AFTER_PROGRAMS,
544 claim: "A CALL that finds no program of its name in the run unit or its program libraries reaches the Language Environment callable service of that name, as a link-edit that finds the name in no user library resolves it from SCEELKED; a user program of the same name comes first",
545 basis: Basis::Chosen,
546 oracle: Oracle::EnterpriseCobol,
547 },
548 Assumption {
549 id: LE_ARGUMENTS_BY_ADDRESS,
550 claim: "A callable service reads and stores each parameter at its argument's address for the length SA38-0683-60 gives it (fullword, halfword-prefixed string, 8-byte COMP-2, 17- or 80-byte string, 12-byte feedback code), whatever the argument's own length; an OMITTED parameter other than fc is a protection exception (S0C4), and a store past the end of the run unit's storage is lost",
551 basis: Basis::Chosen,
552 oracle: Oracle::EnterpriseCobol,
553 },
554 Assumption {
555 id: LE_FEEDBACK_TOKEN,
556 claim: "A feedback code is 12 bytes: severity and message number as halfwords, a byte of case 1, severity and control 001 (X'59' at severity 3), the facility CEE in EBCDIC, then the instance-specific word; twelve zero bytes are CEE000, success (SA38-0683-60, _FEEDBACK in Table 20; so CEE2EB, severity 3 and message 2507, is X'000309CB59C3C5C5')",
557 basis: Basis::Documented,
558 oracle: Oracle::EnterpriseCobol,
559 },
560 Assumption {
561 id: LE_FEEDBACK_NO_INSTANCE_INFO,
562 claim: "The instance-specific word of every feedback code the provided services return is zero",
563 basis: Basis::Chosen,
564 oracle: Oracle::EnterpriseCobol,
565 },
566 Assumption {
567 id: LE_OMITTED_FC_ABENDS,
568 claim: "With fc OMITTED a failing service signals its condition (SA38-0683-60, Invoking callable services); nothing handles it, so one of severity 2 or more ends the run as the default ABTERMENC(ABEND) does, with user abend U4038, and one of severity 1 lets the run continue",
569 basis: Basis::Recalled,
570 oracle: Oracle::EnterpriseCobol,
571 },
572 Assumption {
573 id: LE_CEE3ABD,
574 claim: "CEE3ABD ends the run with user abend abcode modulo 4096, the ABEND macro's user completion code, to which SA38-0683-60 says abcode passes unchecked; every clean-up value ends it alike: open files are closed, as at any U code (C152), and neither a CEEDUMP nor a system dump is written",
575 basis: Basis::Chosen,
576 oracle: Oracle::EnterpriseCobol,
577 },
578 Assumption {
579 id: LE_LOCAL_TIME_IS_UTC,
580 claim: "Local time is UTC: CEELOCT and CEEGMT read the clock that ACCEPT FROM DATE and TIME and CURRENT-DATE read, and CEEGMTO reports a zero offset, as CURRENT-DATE reports +0000",
581 basis: Basis::Chosen,
582 oracle: Oracle::EnterpriseCobol,
583 },
584 Assumption {
585 id: LE_SECONDS_HFP,
586 claim: "Lilian seconds are the whole number of milliseconds converted to long HFP and divided by 1000, truncating as FLOAT_FROM_DECIMAL does, so a MOVE of CEESECS's result to a decimal item can show a millisecond less; seconds given to CEEDATM are taken to the nearest millisecond",
587 basis: Basis::Chosen,
588 oracle: Oracle::EnterpriseCobol,
589 },
590 Assumption {
591 id: LE_PICTURE_OUTPUT,
592 claim: "CEEDATE and CEEDATM write the terms of SA38-0683-60 Table 34 and copy anything else as it stands, with English month and day names; MM after an hour term is minutes (Table 28); CEEDATE writes time terms as zero and AP as AM (Table 27, where Table 35 says blank); the eras <JJJJ>, <CCCC> and YYY are not provided and give CEE2EM; a null or blank picture is the COUNTRY(US) default of Table 33",
593 basis: Basis::Chosen,
594 oracle: Oracle::EnterpriseCobol,
595 },
596 Assumption {
597 id: LE_PICTURE_INPUT,
598 claim: "CEEDAYS and CEESECS read a numeric term followed by a delimiter as up to its width of digits (6/2/88 for MM/DD/YY) and any other as exactly its width, leading blanks allowed; a delimiter takes one character, whatever it is; a month name is three or more of its letters; input that ends before the date is complete is CEE2EB, before the time is complete, zeros; a non-digit is CEE2EO (CEE2ET for CEESECS)",
599 basis: Basis::Chosen,
600 oracle: Oracle::EnterpriseCobol,
601 },
602 Assumption {
603 id: LE_CENTURY_WINDOW,
604 claim: "A two-digit year falls in the hundred years starting 80 years before the current year (SA38-0683-60, CEEDAYS and CEESECS); CEESCEN, which moves the window, is not provided",
605 basis: Basis::Documented,
606 oracle: Oracle::EnterpriseCobol,
607 },
608 Assumption {
609 id: LE_RETURN_CODE_UNCHANGED,
610 claim: "A CALL of a callable service leaves RETURN-CODE as it was; SA38-0683-60 leaves register 15 undefined on return",
611 basis: Basis::Chosen,
612 oracle: Oracle::EnterpriseCobol,
613 },
614 Assumption {
615 id: LE_MESSAGE_AND_DUMP_FILES,
616 claim: "CEEMOUT writes to DD SYSOUT, MSGFILE's default ddname, and CEE3DMP to DD CEEDUMP or the ddname FNAME gives, as UTF-8 text lines, the run's first write replacing the file, or to standard error without the DD; CEE3DMP writes its title, date and time, options and the active programs, not storage or control blocks",
617 basis: Basis::Chosen,
618 oracle: Oracle::EnterpriseCobol,
619 },
620 Assumption {
621 id: LE_HEAP,
622 claim: "CEEGTST gives zeroed run-unit storage on a doubleword from heap 0 only, refuses a request above 256 MiB with CEE0PD, and keeps the storage until the run ends; CEEFRST marks it free, and it is not reused",
623 basis: Basis::Chosen,
624 oracle: Oracle::EnterpriseCobol,
625 },
626 Assumption {
627 id: LE_UNDER_CICS,
628 claim: "In a CICS task, CEE3ABD ends the task as EXEC CICS ABEND CANCEL would, with abcode modulo 4096 as a four-digit decimal ABCODE, so HANDLE ABEND does not catch it; CEEMOUT and CEE3DMP write each line as one item on transient data queue CESE, as MSGFILE and any DD are ignored under CICS (SA38-0683-60), without the terminal, transaction and time prefix CICS may put on a CESE record",
629 basis: Basis::Chosen,
630 oracle: Oracle::EnterpriseCobol,
631 },
632 Assumption {
633 id: REPORT_WRITER_PRECOMPILER,
634 claim: "Enterprise COBOL takes the REPORT SECTION, the FD REPORT clause, INITIATE, GENERATE, TERMINATE, PAGE-COUNTER, LINE-COUNTER, PRINT-SWITCH and USE BEFORE REPORTING only through the COBOL Report Writer Precompiler, 5798-DYR (Migration Guide GC27-8715-04, pp. 69-70); a Report Writer program runs as the precompiler's generated COBOL runs, with the precompiler as supplied, option OSVS on (SC26-4301-04, 1.1.3)",
635 basis: Basis::Documented,
636 oracle: Oracle::EnterpriseCobol,
637 },
638 Assumption {
639 id: REPORT_TOTALS_BEFORE_PAGE_FIT,
640 claim: "Under OSVS a report group's totalling (cross-footing, subtotalling, rolling forward) comes before its USE BEFORE REPORTING procedure and its page-fit test, so a group that forces a new page is already in the totals the PAGE FOOTING shows (SC26-4301-04, 4.2.4)",
641 basis: Basis::Documented,
642 oracle: Oracle::EnterpriseCobol,
643 },
644 Assumption {
645 id: REPORT_SOURCE_SUM_CORRELATION,
646 claim: "Under OSVS a SUM of an item outside the REPORT SECTION that a DETAIL group has as a SOURCE is added only when such a DETAIL is generated, and once for each such DETAIL on GENERATE report-name; UPON names the DETAILs outright; any other such operand is added on every GENERATE (SC26-4301-04, 3.23.5 and 4.2.3)",
647 basis: Basis::Documented,
648 oracle: Oracle::EnterpriseCobol,
649 },
650 Assumption {
651 id: REPORT_PAGE_REGION_DEFAULTS,
652 claim: "With no FOOTING, LAST CONTROL FOOTING is LAST DETAIL when that is written, else the line before the PAGE FOOTING (for a relative PAGE FOOTING, the line that puts its last line on PAGE LIMIT), else PAGE LIMIT; the standard would take PAGE LIMIT. With no LAST DETAIL it is FOOTING; a PAGE LIMIT below either is raised to it (SC26-4301-04, 2.9.3 and message RW-031)",
653 basis: Basis::Documented,
654 oracle: Oracle::EnterpriseCobol,
655 },
656 Assumption {
657 id: REPORT_LINE_WRITES,
658 claim: "Each report line is one WRITE AFTER ADVANCING: the first line of a page AFTER ADVANCING PAGE when it is line 1, otherwise after a record of spaces, with no CODE, written AFTER ADVANCING PAGE; every other line after the distance from the last; a report with no PAGE LIMIT never skips to a new page. The manual shows this for its file handlers and says the direct output also writes line 1 at the top of the page (SC26-4301-04, 5.3.8 and 6.3)",
659 basis: Basis::Chosen,
660 oracle: Oracle::EnterpriseCobol,
661 },
662 Assumption {
663 id: REPORT_CARRIAGE_CONTROL,
664 claim: "A report file is written AFTER ADVANCING, so each report record carries an ASA control character (PRINT_CONTROL_CHARACTER): under ADV a byte before the record, under NOADV the record's first byte, which the precompiler leaves for it (SC26-4301-04, 2.2.3 rules 7 to 9, 2.7.2 rule 3). The CODE comes after the control character, as 5.3.2 says of the PRNT handler, which prints as no handler does, though 2.5.3 rule 2 puts it before",
665 basis: Basis::Chosen,
666 oracle: Oracle::EnterpriseCobol,
667 },
668 Assumption {
669 id: REPORT_RECORD_LENGTH,
670 claim: "A report file whose FD has no RECORD CONTAINS has records as long as the longest line of its reports, rounded up to a multiple of 4, plus the CODE and the control character (SC26-4301-04, 2.2.3 rule 7), the FD's record holding the control character only under NOADV (rule 8, 2.7.2 rule 3), so that under ADV a RECORD CONTAINS length is the line and CODE alone; under RECORDING MODE V each record ends after its last printed field (rule 9)",
671 basis: Basis::Documented,
672 oracle: Oracle::EnterpriseCobol,
673 },
674 Assumption {
675 id: REPORT_SUM_OVERFLOW,
676 claim: "A SUM total is a signed binary item with the integer and decimal places of its SUM entry, widened to those of a REPORT SECTION item it totals, and packed decimal beyond 18 digits (SC26-4301-04, 3.23.4); an addition that would overflow it is not made and run-time error 11 is logged, as 2.8.3 says for SUM OVERFLOW STANDARD, though 3.23.8 says the field then prints blank",
677 basis: Basis::Chosen,
678 oracle: Oracle::EnterpriseCobol,
679 },
680 Assumption {
681 id: REPORT_SOURCE_OVERFLOW,
682 claim: "A SOURCE arithmetic expression that overflows its field or divides by zero leaves the field blank and logs run-time error 10, OVERFLOW PROCEDURE IS STANDARD being the default (SC26-4301-04, 2.8.3)",
683 basis: Basis::Documented,
684 oracle: Oracle::EnterpriseCobol,
685 },
686 Assumption {
687 id: REPORT_SUPPRESS_PRINTING,
688 claim: "SUPPRESS PRINTING, or PRINT-SWITCH left non-zero by a USE BEFORE REPORTING procedure, stops the group's lines, page-fit test and NEXT GROUP, but its totals are still reset; 4.5.3 and 4.7.3 of SC26-4301-04 say so, while 4.2.4 step 9 says no further action is taken",
689 basis: Basis::Chosen,
690 oracle: Oracle::EnterpriseCobol,
691 },
692 Assumption {
693 id: REPORT_NEW_PAGES,
694 claim: "LINE ... NEXT PAGE skips to a new page only when a body group is already on the page; a REPORT FOOTING on a page of its own gets no PAGE HEADING or PAGE FOOTING; a page holding the REPORT HEADING alone gets no PAGE FOOTING (SC26-4301-04, 3.24.3)",
695 basis: Basis::Chosen,
696 oracle: Oracle::EnterpriseCobol,
697 },
698 Assumption {
699 id: REPORT_OUT_OF_ORDER,
700 claim: "GENERATE for a report not initiated logs run-time error 14 and initiates it (SC26-4301-04, message RW-142); INITIATE of an active report starts it afresh; TERMINATE of an inactive report does nothing",
701 basis: Basis::Chosen,
702 oracle: Oracle::EnterpriseCobol,
703 },
704 Assumption {
705 id: REPORT_CONTROL_AREA,
706 claim: "PAGE-COUNTER, LINE-COUNTER and PRINT-SWITCH are S9(9) COMP items (PAGE-COUNTER's PICTURE is in SC26-4301-04, 3.15.2); each report's control area, fields and totals are WORKING-STORAGE after the program's own items, so CANCEL and IS INITIAL start the report afresh",
707 basis: Basis::Chosen,
708 oracle: Oracle::EnterpriseCobol,
709 },
710 Assumption {
711 id: OBJECT_REFERENCE_VALUE,
712 claim: "An object reference is four bytes, as under LP(32): zero for NULL, otherwise the number of the local or global reference it holds, never reused in the run unit; each NEW, RETURNING value, argument received, SELF and reference a JNI service makes is a new number, so two references to one object can differ byte for byte, and = compares the objects they identify (Language Reference SC27-8713-03, p. 282)",
713 basis: Basis::Chosen,
714 oracle: Oracle::EnterpriseCobol,
715 },
716 Assumption {
717 id: LOCAL_REFERENCES_EXPIRE,
718 claim: "The object references a method receives as arguments, gets back as INVOKE RETURNING values or from JNI services, and makes with INVOKE ... NEW are local references, valid until the method returns, whatever kind the invoked method returned; NewGlobalRef makes a global reference, valid until DeleteGlobalRef, and DeleteLocalRef frees a local reference at once (Programming Guide SC27-8714-03, pp. 702-703 and 721-723; Language Reference SC27-8713-03, p. 365). SET copies a reference and converts nothing (Language Reference p. 451), so the Guide's 'use a SET statement to convert' (p. 722) is read as the CALL of NewGlobalRef it shows on p. 702, and a local reference kept in OBJECT, FACTORY or method WORKING-STORAGE expires, which p. 722 calls an error",
719 basis: Basis::Documented,
720 oracle: Oracle::EnterpriseCobol,
721 },
722 Assumption {
723 id: OBJECTS_NEVER_FREED,
724 claim: "Objects are never freed, and a run unit that creates more than 1,000,000 of them abends, where Java's garbage collector reclaims objects no longer referred to",
725 basis: Basis::Chosen,
726 oracle: Oracle::EnterpriseCobol,
727 },
728 Assumption {
729 id: INSTANCE_DATA_START,
730 claim: "INVOKE class NEW gives each COBOL class in the new object's hierarchy its own copy of its OBJECT WORKING-STORAGE, set to X'00' and then to its VALUE clauses; IBM documents only the VALUE clauses",
731 basis: Basis::Chosen,
732 oracle: Oracle::EnterpriseCobol,
733 },
734 Assumption {
735 id: FACTORY_DATA_START,
736 claim: "A class's FACTORY WORKING-STORAGE is set to X'00' and then to its VALUE clauses when the run unit first uses the class, and its one copy serves every INVOKE of the class's factory methods, through a subclass too",
737 basis: Basis::Chosen,
738 oracle: Oracle::EnterpriseCobol,
739 },
740 Assumption {
741 id: METHOD_LOOKUP,
742 claim: "INVOKE selects the method by its name, case kept, and the Java types of its arguments and RETURNING item (void without one), a universal object reference counting as java.lang.Object; the search starts at the class of the object itself, not the class its reference is typed with, and goes up the INHERITS chain; SUPER starts at the parent of the class that defines the running method",
743 basis: Basis::Chosen,
744 oracle: Oracle::EnterpriseCobol,
745 },
746 Assumption {
747 id: METHOD_NAME_ITEM,
748 claim: "A method name held in a data item is the item's content without its trailing spaces",
749 basis: Basis::Chosen,
750 oracle: Oracle::EnterpriseCobol,
751 },
752 Assumption {
753 id: INVOKE_NULL,
754 claim: "INVOKE on a NULL object reference, or on four bytes that name no object, ends the run with an abend; IBM leaves the result undefined",
755 basis: Basis::Chosen,
756 oracle: Oracle::EnterpriseCobol,
757 },
758 Assumption {
759 id: NO_METHOD_ABEND,
760 claim: "An INVOKE without ON EXCEPTION that finds no method raises IBM's severity-3 Language Environment condition, which ends the run with abend U4038",
761 basis: Basis::Chosen,
762 oracle: Oracle::EnterpriseCobol,
763 },
764 Assumption {
765 id: FACTORY_SELF,
766 claim: "SELF in a factory method is the factory object of the class that defines the method, and INVOKE on a reference to a factory object runs factory methods",
767 basis: Basis::Chosen,
768 oracle: Oracle::EnterpriseCobol,
769 },
770 Assumption {
771 id: CLASS_SEARCH,
772 claim: "A class is a COBOL class when a class definition with its external name is among the programs read or in the program libraries, as a member named with its simple name or its full name with periods as underscores, as IBM names the class's DLL; any other class is a Java class",
773 basis: Basis::Chosen,
774 oracle: Oracle::EnterpriseCobol,
775 },
776 Assumption {
777 id: INVOKE_KEEPS_RETURN_CODE,
778 claim: "INVOKE leaves the invoking program's RETURN-CODE as it was (Language Reference, INVOKE statement, RETURNING phrase)",
779 basis: Basis::Documented,
780 oracle: Oracle::EnterpriseCobol,
781 },
782 Assumption {
783 id: OO_OPTIONS_REQUIRED,
784 claim: "A class definition, and a program with INVOKE or an object reference, is compiled with THREAD, DLL, RENT and DBCS (Programming Guide SC27-8714-03, pp. 291, 295, 363, 588, 591, 694; Language Reference SC27-8713-03, p. 89), and NORENT with THREAD or DLL is a conflict IBM resolves as RENT (Guide pp. 344-345). Under THREAD a program is RECURSIVE ('an error will occur', Guide p. 591; Language Reference p. 103), and INITIAL, nested programs, SORT of a file and MERGE are diagnosed as errors (Guide p. 418; Language Reference pp. 85, 103, 400, 452); a table SORT is allowed (p. 453), and a method is recursive without it (p. 94). The options are the CBL and PROCESS cards' over IBM's defaults NOTHREAD, NODLL, RENT and DBCS",
785 basis: Basis::Documented,
786 oracle: Oracle::EnterpriseCobol,
787 },
788 Assumption {
789 id: CHAR_FROM_DISPLAY,
790 claim: "A one-byte reference modification of a display item passed to INVOKE becomes a Java char through the program's code page",
791 basis: Basis::Chosen,
792 oracle: Oracle::EnterpriseCobol,
793 },
794 Assumption {
795 id: SORT_EQUAL_KEYS_IN_ORDER,
796 claim: "Records with equal keys leave a SORT in the order they entered it whether or not WITH DUPLICATES IN ORDER is written, as under DFSORT's EQUALS; a table SORT keeps equal elements in their order too",
797 basis: Basis::Chosen,
798 oracle: Oracle::EnterpriseCobol,
799 },
800 Assumption {
801 id: MERGE_EQUAL_KEYS_BY_FILE,
802 claim: "Records with equal keys leave a MERGE in the order of the USING files, and each file's in its own order",
803 basis: Basis::Chosen,
804 oracle: Oracle::EnterpriseCobol,
805 },
806 Assumption {
807 id: MERGE_OUT_OF_SEQUENCE_FAILS,
808 claim: "A MERGE input file whose records are out of the merge order makes the MERGE fail with SORT-RETURN 16 before any record is output, as DFSORT's ICE068A ends a merge",
809 basis: Basis::Recalled,
810 oracle: Oracle::EnterpriseCobol,
811 },
812 Assumption {
813 id: SORT_FILE_FAILURE,
814 claim: "An I/O failure on a USING or GIVING file with a FILE STATUS sets the status and makes the SORT or MERGE fail with SORT-RETURN 16, leaving the rest undone; without a FILE STATUS it ends the run as the same failure on OPEN, READ, WRITE or CLOSE does",
815 basis: Basis::Chosen,
816 oracle: Oracle::EnterpriseCobol,
817 },
818 Assumption {
819 id: SORT_DECIMAL_KEYS,
820 claim: "A file SORT or MERGE compares a zoned or packed key as DFSORT compares ZD, PD, CLO, CSL and CST fields (z/OS DFSORT Application Programming Guide SC23-6878-50, Appendix C, pp. 825-828): sign nibbles F, E, C, A, 8, 6, 4, 2 and 0 are positive and D, B, 9, 7, 5, 3 and 1 negative, a separate sign is negative only when it is '-', the zones of the other digits are ignored, and no key is a data exception; a binary key compares at its full width (BI, FI) and a floating-point one by value (FL). -strict-sort-keys reads the key as the program would, so an invalid one is S0C7; a table SORT always compares as a relation condition does",
821 basis: Basis::Documented,
822 oracle: Oracle::EnterpriseCobol,
823 },
824 Assumption {
825 id: SORT_RECORD_LENGTHS,
826 claim: "A record shorter than a fixed-length SD or GIVING record is padded with spaces and a longer one cut to that length; a variable-length record that ends inside a key makes the SORT or MERGE fail with SORT-RETURN 16, as DFSORT does without VLSHRT",
827 basis: Basis::Chosen,
828 oracle: Oracle::EnterpriseCobol,
829 },
830 Assumption {
831 id: SORT_RETURN_STOPS,
832 claim: "SORT-RETURN is 0 when a SORT or MERGE starts; 16 moved to it in an input or output procedure stops the operation at the next RELEASE or RETURN, which does nothing, or when the input procedure ends",
833 basis: Basis::Chosen,
834 oracle: Oracle::EnterpriseCobol,
835 },
836 Assumption {
837 id: RETURN_AFTER_END,
838 claim: "A RETURN after the AT END condition takes the AT END phrase again",
839 basis: Basis::Chosen,
840 oracle: Oracle::EnterpriseCobol,
841 },
842 Assumption {
843 id: SORT_KEY_INVALID_DIGITS,
844 claim: "A digit nibble A to F in a zoned or packed sort key collates above 9 in its place; DFSORT does not say where such a key collates",
845 basis: Basis::Chosen,
846 oracle: Oracle::EnterpriseCobol,
847 },
848 Assumption {
849 id: SORT_NEGATIVE_ZERO,
850 claim: "A zoned or packed sort key of -0 collates before +0 in ascending order, as under DFSORT's SZERO=YES (Installation and Customization SC23-6881-70, p. 99), the IBM-supplied default",
851 basis: Basis::Recalled,
852 oracle: Oracle::EnterpriseCobol,
853 },
854 Assumption {
855 id: FASTSRT_FILES,
856 claim: "Under FASTSRT DFSORT does the I/O of a SORT's only USING file and its only GIVING file, except a MERGE's, a line-sequential or variable-length relative file, a GIVING file whose FD has LINAGE (p. 233), a print file under ADV unless --fastsrt-adv-print=include, one whose records differ from the SD's in format (fixed or variable) or largest length, and a GIVING file that is also the USING file; COBOL does the rest as under NOFASTSRT (Programming Guide SC27-8714-03, pp. 232-233, 369). The Guide does not name a print file under ADV. By default, --fastsrt-adv-print=exclude, it is left to COBOL, because ADV adds a byte to its record length for the printer control character (p. 346), which the DD's LRECL counts (p. 185): p. 233 wants the SD's and the FD's largest records the same length and p. 232 the DD to match the FD, so with the FD's record as long as the SD's the data set's records are a byte longer than DFSORT's, and with it a byte shorter the FD's and the SD's differ. --fastsrt-adv-print=include reads p. 233 as the FD's length alone and gives the file to DFSORT, as FASTSRT_ADV_PRINT says. Under NOADV the character is inside the FD's record and the lengths rule alone applies",
857 basis: Basis::Documented,
858 oracle: Oracle::EnterpriseCobol,
859 },
860 Assumption {
861 id: FASTSRT_STATUS,
862 claim: "A file whose I/O DFSORT does keeps the FILE STATUS it had through the SORT, and a GIVING relative file's RELATIVE KEY is not set (Programming Guide SC27-8714-03, pp. 232-233)",
863 basis: Basis::Documented,
864 oracle: Oracle::EnterpriseCobol,
865 },
866 Assumption {
867 id: FASTSRT_FAILURE,
868 claim: "A failure on a file whose I/O DFSORT does, including an empty VSAM input file, fails the SORT with SORT-RETURN 16 and the run goes on, with or without a FILE STATUS",
869 basis: Basis::Chosen,
870 oracle: Oracle::EnterpriseCobol,
871 },
872 Assumption {
873 id: SAME_AREA_VSAM,
874 claim: "SAME AREA makes the VSAM (indexed and relative) files it names share one record area, as SAME RECORD AREA does, and is documentation for the others (Language Reference SC27-8713-03, p. 156)",
875 basis: Basis::Documented,
876 oracle: Oracle::EnterpriseCobol,
877 },
878 Assumption {
879 id: LONG_ZONED_BY_PACKS,
880 claim: "A zoned item of more than 16 digits, too long for one PACK (SA22-7832-14: each operand at most 16 bytes), enters arithmetic through two or three PACKs, the high-order part first and each lower PACK overwriting the byte the part above it ended in, so it packs as ZONED_BY_PACK says one PACK would: every zone but the sign's discarded and no digit checked until the arithmetic uses it. An alphanumeric sender longer than 31 characters moved to a numeric item is packed from its rightmost 31, the most any receiver holds",
881 basis: Basis::Chosen,
882 oracle: Oracle::EnterpriseCobol,
883 },
884 Assumption {
885 id: ASCII_COLLATION,
886 claim: "STANDARD-1 and STANDARD-2 put the characters of 7-bit ASCII in its order (Language Reference SC27-8713-03, Table 82, p. 754; Programming Guide SC27-8714-03, p. 7), each found in the program's code page; the characters that are not 7-bit ASCII follow them in EBCDIC order, as the characters an ALPHABET literal leaves out do",
887 basis: Basis::Chosen,
888 oracle: Oracle::EnterpriseCobol,
889 },
890 Assumption {
891 id: TABLE_SORT_COLLATION,
892 claim: "A table SORT without a COLLATING SEQUENCE phrase orders alphanumeric keys in EBCDIC, as the SORT statement's format 2 rules say (SC27-8713-03, p. 450), though its rules for the phrase in both formats put the PROGRAM COLLATING SEQUENCE in the phrase's place (p. 451)",
893 basis: Basis::Chosen,
894 oracle: Oracle::EnterpriseCobol,
895 },
896 Assumption {
897 id: ALPHABET_LITERALS,
898 claim: "Among an ALPHABET clause's literals HIGH-VALUE, LOW-VALUE, SPACE, ZERO and QUOTE are the EBCDIC characters X'FF', X'00', X'40', X'F0' and X'7F', whatever the alphabet makes HIGH-VALUE and LOW-VALUE, and a numeric literal n is the character at ordinal n of EBCDIC, whose ordinals SC27-8713-03 points to (p. 128)",
899 basis: Basis::Chosen,
900 oracle: Oracle::EnterpriseCobol,
901 },
902 Assumption {
903 id: LE_CEEIGZCT,
904 claim: "A COPY CEEIGZCT that no library answers gives a level-88 condition name for each of the 723 conditions whose entry in the Language Environment Runtime Messages (SA38-0686-60, chapter 1) shows a symbolic feedback code, severity from its message's I, W, E, S or C; the name is CEE and the message number in base 32 (Programming Guide SA38-0682-60, CEEBLDTX :msgname.), the value the token's first 8 bytes as LE_FEEDBACK_TOKEN lays them out (Programming Reference SA38-0683-60, CEENCOD), and CEE000 is all zeros (SA38-0682-60, testing a condition token for success); it is meant to follow the 8-byte group the token starts with (SA38-0682-60, Figure 76). It is written from those manuals alone, not from IBM's CEEIGZCT",
905 basis: Basis::Documented,
906 oracle: Oracle::EnterpriseCobol,
907 },
908 Assumption {
909 id: LE_CEEIGZCT_DISAGREEMENTS,
910 claim: "Where the manuals disagree CEEIGZCT follows the message: its severity letter over a service's table in SA38-0683-60 (CEE07V, CEE317, CEE35S, CEE36V to CEE374), and its number over the code SA38-0686-60 prints beside it (CEE0356C shows CEE0BA, CEE5722I CEE5IP, CEE5771S CEE5KC); the eight messages printed with no code (CEE3252E, CEE3257E to CEE3259E, CEE3596S, CEE3796I to CEE3798I) have no name",
911 basis: Basis::Chosen,
912 oracle: Oracle::EnterpriseCobol,
913 },
914 Assumption {
915 id: LE_SHORT_ARGUMENT_LIST,
916 claim: "A CALL of a service with fewer arguments than its syntax lists, CEE3ABD with no USING among them, ends the run with ironwork's own abend, not a modelled one: IBM calls a short list invalid with unpredictable results (SA38-0683-60, General usage notes for callable services) and says nothing of register 1 at a CALL without USING, whose own CALL and CEEPCALL macros leave it unaltered when no parameter is coded (MVS Assembler Services Reference SA22-7606-13, CALL; SA38-0682-60, CEEPCALL); the service then reads its arguments through whatever register 1 and the storage past the list hold, so neither S0C4 nor any other result follows. A missing fc is not taken as OMITTED, nor a missing clean-up as none",
917 basis: Basis::Chosen,
918 oracle: Oracle::EnterpriseCobol,
919 },
920 Assumption {
921 id: LE_SSRANGE_U4038,
922 claim: "A reference SSRANGE finds out of range signals a severity-3 LE condition; nothing handles it, so the run ends with user abend U4038 under the default ABTERMENC(ABEND). The message ids, from memory of the Enterprise COBOL messages: IGZ0006S for a subscript or index outside its table, IGZ0072S for a reference-modification start outside its item, IGZ0073S for a length below 1, and IGZ0074S for a start and length that reach past the item's end. An OCCURS DEPENDING ON object outside its range gives IGZ0007S, IBM's message for a variable-length group longer than its maximum or shorter than zero (IBM Support, 'Executing a CICS generated application results in abend'); IBM checks the group's composite length, and ironwork checks each object's count",
923 basis: Basis::Recalled,
924 oracle: Oracle::EnterpriseCobol,
925 },
926 Assumption {
927 id: PRINT_CONTROL_CHARACTER,
928 claim: "A sequential file that a WRITE with ADVANCING in the program names, or whose FD has LINAGE, is a print file: every record written to it carries a printer control character, a WRITE without ADVANCING being AFTER ADVANCING 1 LINE; ASA characters when every WRITE ... ADVANCING of the file says AFTER, machine codes when any says BEFORE (Language Reference SC27-8713-03, p. 479). Under ADV, the default, the character is a byte before the record; under NOADV it is the record's own first byte; a LINAGE file is ADV whatever the option (p. 480; Programming Guide SC27-8714-03, pp. 178-179, 346). ASA ' ', '0' and '-' space 1 to 3 lines before printing, '+' none, '1' to '9' and 'A' to 'C' skip to channels 1 to 12, PAGE and C01 being channel 1, except that PAGE moves a LINAGE file's paper in lines (LINAGE_PAGE_MOVEMENT), and CSP '+'; machine codes print then space, X'01', X'09', X'11', X'19', or skip, X'89' + 8(n-1) for channel n; AFP-5A is X'5A' (z/OS DFSMS Macro Instructions for Data Sets SC23-6852-60, pp. 397-398; Language Reference pp. 126-127, 483). SPECIAL-NAMES of a program apply to the programs it contains (p. 13)",
929 basis: Basis::Documented,
930 oracle: Oracle::EnterpriseCobol,
931 },
932 Assumption {
933 id: PRINT_SPACING_RECORDS,
934 claim: "A movement one control character cannot give is made with records that only move the paper, each as long as the line's record and blank after its control character: in an ASA file AFTER ADVANCING n lines above 3 is preceded by (n-1)/3 records with '-', the line taking the rest; in a machine-code file BEFORE ADVANCING n above 3 prints with X'19' and is followed by records spacing the rest without printing, three lines at a time (X'1B', then X'0B' or X'13'), and AFTER ADVANCING is such records, or X'8B' + 8(n-1) for channel n, then the line with X'01'",
935 basis: Basis::Chosen,
936 oracle: Oracle::EnterpriseCobol,
937 },
938 Assumption {
939 id: PRINT_CONTROL_RUN_TIME,
940 claim: "Under NOADV the control character is stored in the first byte of the record area, where the program sees it after the WRITE; an ADVANCING count below zero spaces no lines; under ADV a READ of a print file in the program that writes it skips the added byte",
941 basis: Basis::Chosen,
942 oracle: Oracle::EnterpriseCobol,
943 },
944 Assumption {
945 id: TEXT_PRINT_LINES,
946 claim: "A text DD shows a print file's records without the control character, which it reads as line spacing, as the POSIX asa utility does: ' ', '0' and '-' put the line one, two or three lines below the last, the DD's first line starting on its first; '+' and X'01' overprint, after a carriage return when both lines show something; a skip to channel 1 is a form feed before the line; a skip to channels 2 to 12, or AFP-5A page mode data, is one line, a text DD having no forms control buffer; a machine code moves the paper after its line. Any other file's records are a line each",
947 basis: Basis::Chosen,
948 oracle: Oracle::EnterpriseCobol,
949 },
950 Assumption {
951 id: LOCAL_REFERENCES_OUTSIDE_METHODS,
952 claim: "A local reference made by a program that is not a method belongs to the method running when it runs, freed when that method returns; with no method running it belongs to the run unit and stays valid until the run ends, as the JNI keeps a thread's local references outside any native method, and the Guide asks for NewGlobalRef only 'if the client code is within a method' (Programming Guide SC27-8714-03, p. 703)",
953 basis: Basis::Chosen,
954 oracle: Oracle::EnterpriseCobol,
955 },
956 Assumption {
957 id: SELF_IS_LOCAL,
958 claim: "SELF is a local reference of the method's own frame, made with the invocation, as the JNI passes a native method its object, and it expires when the method returns; the Guide's list of local references names arguments, RETURNING values, JNI results and NEW only (Programming Guide SC27-8714-03, p. 721)",
959 basis: Basis::Recalled,
960 oracle: Oracle::EnterpriseCobol,
961 },
962 Assumption {
963 id: EXPIRED_REFERENCE_ABENDS,
964 claim: "Using a reference after it was freed ends the run with abend IRONWORK naming the item that holds it, how the reference was made and where it was freed: INVOKE on it, passing it to INVOKE or a JNI service, a method's RETURNING it, and comparing it with another object reference or SELF use it; SET, MOVE and CALL between programs copy its bytes without looking, and a comparison with NULL tests only its bytes. On z/OS a freed reference's slot is reused, and the Guide says only that 'an error occurs' (Programming Guide SC27-8714-03, p. 722)",
965 basis: Basis::Chosen,
966 oracle: Oracle::EnterpriseCobol,
967 },
968 Assumption {
969 id: LOCAL_FRAMES,
970 claim: "The JNI reference services run as the JNI specification defines them: NewLocalRef and NewGlobalRef make a new reference to the object, NULL for NULL; DeleteLocalRef and DeleteGlobalRef free one and ignore NULL, and one given the other kind of reference ends the run; IsSameObject compares objects; GetObjectRefType answers 0 for NULL, 1 for local and 2 for global; PushLocalFrame opens a frame whose local references PopLocalFrame frees, giving back a local reference in the frame below to its argument's object, and EnsureLocalCapacity succeeds. PopLocalFrame with no frame of PushLocalFrame's open ends the run, and a method's return frees the frames it left open. The Guide documents NewGlobalRef, DeleteGlobalRef and DeleteLocalRef only (Programming Guide SC27-8714-03, pp. 722-723)",
971 basis: Basis::Recalled,
972 oracle: Oracle::EnterpriseCobol,
973 },
974 Assumption {
975 id: OO_OPTIONS_SEVERITY,
976 claim: "IBM names no message for object-oriented syntax compiled without THREAD, DLL, RENT or DBCS, nor for THREAD without RECURSIVE, which it calls an error (Messages and Codes SC27-4648-02, p. v, lists only some messages), so the severity of each rule of J13 is chosen. A missing option is a warning (W, return code 4), and the program runs as if compiled with it. NORENT with THREAD or DLL is a warning too: IBM forces RENT and 'generates an error message' (Programming Guide SC27-8714-03, p. 344), and the message it gives for an option dropped in conflict resolution is W, IGYOS4020-W, return code 4, in Enterprise COBOL job output quoted in the corpus (Delvoie_Mainframe, A5 instructions). THREAD without RECURSIVE, and INITIAL, a nested program, SORT of a file or MERGE under THREAD, are errors (S), as IBM diagnoses them as errors (J13). A program that reaches Java through JNIENVPTR alone, with no INVOKE or object reference, is held to none of them, as IBM builds its Bank-of-Z IBTRAN with DLL and without THREAD",
977 basis: Basis::Chosen,
978 oracle: Oracle::EnterpriseCobol,
979 },
980 Assumption {
981 id: REFERENCES_KEPT,
982 claim: "Every reference made is kept for the rest of the run, so that an expired one can say where it expired, and a run unit that makes more than 8,388,608 of them abends; the JVM reuses a freed reference's slot",
983 basis: Basis::Chosen,
984 oracle: Oracle::EnterpriseCobol,
985 },
986 Assumption {
987 id: SQL_COMMIT_AT_NORMAL_END,
988 claim: "A batch run unit commits at a normal end and rolls back at an abend: \"In all Db2 environments, the normal termination of a process is an implicit commit operation\" (Db2 12 for z/OS SQL Reference, COMMIT). Db2 for Linux rolls back instead",
989 basis: Basis::Documented,
990 oracle: Oracle::Db2,
991 },
992 Assumption {
993 id: SQL_WHENEVER_ORDER,
994 claim: "WHENEVER tests SQLERROR (SQLCODE < 0), NOT FOUND (100) and SQLWARNING (SQLWARN0 W, or > 0 and not 100); the precompiler tests warning before not-found, and the three exclude one another (Db2 12.1.5 for Linux, tools/db2-probe)",
995 basis: Basis::Observed,
996 oracle: Oracle::Db2,
997 },
998 Assumption {
999 id: SQL_TRUNCATED_INDICATOR,
1000 claim: "A string cut to fit its host variable sets the indicator to its original length, SQLWARN0 and SQLWARN1, and SQLSTATE 01004; trailing blanks cut from a CHAR count (Db2 12.1.5 for Linux, tools/db2-probe)",
1001 basis: Basis::Observed,
1002 oracle: Oracle::Db2,
1003 },
1004 Assumption {
1005 id: SQL_POSTGRES_ERRORS,
1006 claim: "PostgreSQL's SQLSTATEs map to Db2 SQLCODEs as sql-runtime.md §9 tabulates; Db2 for Linux agrees except -433 for -404 and -801 for -802, where the table keeps z/OS's documented codes (Db2 12.1.5 for Linux, tools/db2-probe)",
1007 basis: Basis::Observed,
1008 oracle: Oracle::Db2,
1009 },
1010 Assumption {
1011 id: SQL_DIALECT_REWRITES,
1012 claim: "Db2 SQL is rewritten for PostgreSQL by the table in sql-runtime.md §9, and any other text runs unchanged",
1013 basis: Basis::Chosen,
1014 oracle: Oracle::Db2,
1015 },
1016 Assumption {
1017 id: SQL_INTO_WITHOUT_COLONS,
1018 claim: "A name in an INTO list written without its colon is a host variable, as older precompilers read it",
1019 basis: Basis::Recalled,
1020 oracle: Oracle::Db2,
1021 },
1022 Assumption {
1023 id: SQL_DECLARATIONS_CROSS_NESTED_PROGRAMS,
1024 claim: "WHENEVER and cursor declarations carry on in listing order across nested programs, as the precompiler reads the source in order",
1025 basis: Basis::Chosen,
1026 oracle: Oracle::Db2,
1027 },
1028 Assumption {
1029 id: SQL_DOUBLE_TO_HFP_TRUNCATES,
1030 claim: "An IEEE double stored into COMP-1 or COMP-2 drops the low-order bits that do not fit, rather than rounding",
1031 basis: Basis::Chosen,
1032 oracle: Oracle::Db2,
1033 },
1034 Assumption {
1035 id: SQL_ZONED_IS_DECIMAL,
1036 claim: "A zoned DISPLAY host variable without SIGN SEPARATE is DECIMAL to Db2, as SIGN LEADING SEPARATE is",
1037 basis: Basis::Chosen,
1038 oracle: Oracle::Db2,
1039 },
1040 Assumption {
1041 id: SQL_ISO_DATETIME,
1042 claim: "Dates and times reach character host variables as YYYY-MM-DD, HH.MM.SS and YYYY-MM-DD-HH.MM.SS.NNNNNN, DSNHDECP's DATE(ISO) and TIME(ISO); the forms are as seen under DATETIME(ISO), and the default is an installation's (Db2 12.1.5 for Linux, tools/db2-probe)",
1043 basis: Basis::Observed,
1044 oracle: Oracle::Db2,
1045 },
1046 Assumption {
1047 id: SQL_TRAILING_BLANKS_SENT,
1048 claim: "Character inputs are sent with their trailing blanks, since Db2 compares strings as if blank-padded (Db2 12.1.5 for Linux, tools/db2-probe)",
1049 basis: Basis::Observed,
1050 oracle: Oracle::Db2,
1051 },
1052 Assumption {
1053 id: SQL_FETCH_ROW_COUNT,
1054 claim: "A single-row FETCH that returns a row sets SQLERRD(3) to 1; Db2 for z/OS documents SQLERRD(3) for a rowset FETCH only (Db2 12.1.5 for Linux, tools/db2-probe)",
1055 basis: Basis::Observed,
1056 oracle: Oracle::Db2,
1057 },
1058 Assumption {
1059 id: FASTSRT_PRINT_RECORDS,
1060 claim: "Under FASTSRT DFSORT, not COBOL, does the I/O of the USING and GIVING files (Programming Guide SC27-8714-03, p. 232), and DFSORT adds a printer control character only to the lines of an OUTFIL report (DFSORT Application Programming Guide SC23-6878-50, p. 225), which needs control statements ironwork does not read. So DFSORT writes a GIVING print file's records as the SD holds them, with no control character, where COBOL writes each as a WRITE without phrases (Language Reference SC27-8713-03, p. 453), which for a print file is AFTER ADVANCING 1 LINE (Programming Guide pp. 178-179), the character over the record's first byte under NOADV; and DFSORT reads a USING print file's records as its data set holds them, where COBOL's READ skips the byte ADV adds (PRINT_CONTROL_RUN_TIME). Under NOADV the two read a USING print file alike",
1061 basis: Basis::Documented,
1062 oracle: Oracle::EnterpriseCobol,
1063 },
1064 Assumption {
1065 id: FASTSRT_RECORD_LENGTHS,
1066 claim: "DFSORT takes the length of the records it sorts from the SORTIN data set, not from the RECORD statement (DFSORT Application Programming Guide SC23-6878-50, p. 420). A fixed-length record shorter than SORTOUT's LRECL is padded on the right with X'00', and a longer one cut on the right, with message ICE171I and return code 0 (pp. 14-15, 200, 209) under PAD=RC0 and TRUNC=RC0, the IBM-supplied defaults (DFSORT Installation and Customization SC23-6881-70, pp. 91, 102); padding needs the Blockset technique and a sort or copy, DFSORT checks neither padding nor truncation without both a SORTIN and a SORTOUT data set (pp. 200, 209), and it neither pads nor cuts the records an E15 or E35 exit returns (p. 15). A variable-length record longer than SORTOUT's LRECL ends DFSORT with ICE217A under NOVLLONG (p. 210), the IBM-supplied default (Installation and Customization p. 103; DFSORT Messages, Codes and Diagnosis SC23-6879-50, p. 76)",
1067 basis: Basis::Documented,
1068 oracle: Oracle::EnterpriseCobol,
1069 },
1070 Assumption {
1071 id: FASTSRT_ADV_PRINT,
1072 claim: "With --fastsrt-adv-print=include a print file under ADV is FASTSRT's as any other file is, and DFSORT meets its data set's records, a byte longer than the FD's. A USING file's data set stands as DFSORT's SORTIN and a GIVING file's as its SORTOUT, as the Programming Guide implies by keeping DFSORT's SORTIN and SORTOUT options from a FASTSRT program (SC27-8714-03, p. 232) without saying so, and COBOL gives DFSORT each key's place in the SD's record; so, by FASTSRT_PRINT_RECORDS and FASTSRT_RECORD_LENGTHS, a USING print file's records keep the control character as their first byte, each key being read a byte before where the FD has it, and a GIVING file's fixed-length records are padded with X'00' or cut to its data set's length. With no USING data set of DFSORT's own, as with an INPUT PROCEDURE or a USING file COBOL reads, a GIVING print file's longer fixed-length records fail the SORT before its input phase, as ICE043A reason 9 says of fixed-length output records longer than the input's (DFSORT Messages, Codes and Diagnosis SC23-6879-50, pp. 29-30). COBOL takes a record from DFSORT, for an OUTPUT PROCEDURE or a GIVING file it writes, at the SD's length at most, and a VSAM GIVING file takes it at its own. A text DD holds no control character, so on it a print file's records are the FD's length, and each record DFSORT writes is a line",
1073 basis: Basis::Chosen,
1074 oracle: Oracle::EnterpriseCobol,
1075 },
1076 Assumption {
1077 id: COMPILER_SEVERITIES,
1078 claim: "A compiler message has one of five severities, each with a return code: I (informational) 0, the program runs correctly; W (warning) 4, a possible error; E (error) 8, an error the compiler attempted to correct; S (severe) 12, one it could not, and the program should not be run; U (unrecoverable) 16, the compilation ended. A compilation's return code is generally the highest of its messages' (Programming Guide SC27-8714-03, Table 38, p. 282), and the letter ends each message's identifier, as in IGYPS2121-S (p. 281). ironwork check exits with that return code, 0 when there is no message",
1079 basis: Basis::Documented,
1080 oracle: Oracle::EnterpriseCobol,
1081 },
1082 Assumption {
1083 id: REFUSALS_ARE_SEVERE,
1084 claim: "Every refusal ironwork made before its messages had severities is S, return code 12, and none is E: IBM's E is an error the compiler corrects, still producing object code under the default NOCOMPILE(S) (Programming Guide SC27-8714-03, pp. 282, 355), and ironwork corrects nothing, so a program it refuses has no code to run. Where IBM documents a lower severity ironwork now follows it: NUMPROC(MIG) is W with the default NUMPROC, an invalid suboption is E with the option discarded, and a non-COBOL character is IGYLI0163-E (C120 onward). U is not used: ironwork's reader stops at the first syntax error, which IBM reports and reads past",
1085 basis: Basis::Chosen,
1086 oracle: Oracle::EnterpriseCobol,
1087 },
1088 Assumption {
1089 id: REFUSED_FROM_S,
1090 claim: "ironwork run and cics run a program whose compile's return code is 0, 4 or 8, printing its messages first, and refuse one at 12 or more, exiting with that return code, as IBM's IGYWCLG procedure runs one: its GO step is bypassed only when 8 is less than the compile step's return code, COND=((8,LT,COBOL),(4,LT,LKED)) (Programming Guide SC27-8714-03, pp. 259-260), and the default NOCOMPILE(S) produces object code after E-level messages, stopping it at the first S-level one (p. 355). A card's COMPILE or NOCOMPILE, or -warnings-block, moves the refusal (NOCOMPILE). No message of ironwork's is E yet (REFUSALS_ARE_SEVERE)",
1091 basis: Basis::Documented,
1092 oracle: Oracle::EnterpriseCobol,
1093 },
1094 Assumption {
1095 id: NOCOMPILE,
1096 claim: "COMPILE, abbreviated C, produces object code whatever the messages; NOCOMPILE(W), NOCOMPILE(E) or NOCOMPILE(S), abbreviated NOC, stops it at the first message of that severity or higher, NOCOMPILE(S) being the default; and NOCOMPILE alone is a syntax check with no object code (Programming Guide SC27-8714-03, p. 355). ironwork reads them from a CBL or PROCESS card, the last one given winning (p. 344): run and cics refuse a program with a message at or above the level, under NOCOMPILE whatever its messages, and under COMPILE from S as under NOCOMPILE(S): COMPILE's object code after an S-level message runs with results IBM calls unpredictable (p. 355), and IGYWCLG's COND bypasses its GO step above 8 whatever the object code (REFUSED_FROM_S). A program ironwork cannot parse or lay out is refused under any of them. -warnings-block is ironwork's command-line NOCOMPILE(W), and a card's COMPILE or NOCOMPILE wins over it, as options on a PROCESS or CBL statement take precedence over the compiler invocation's (p. 273). None of them changes the return code, the highest of the messages' (p. 282), so check exits as it would without them. IBM has no option that turns a warning into an error: FLAG(x,y) chooses only which messages are listed (pp. 369-370), and a MSGEXIT user exit of the EXIT option can raise a W or I message to any severity up to S, one message at a time, which changes the return code (pp. 836-837)",
1097 basis: Basis::Chosen,
1098 oracle: Oracle::EnterpriseCobol,
1099 },
1100 Assumption {
1101 id: COMMENT_ENTRY_EXTENT,
1102 claim: "The comment-entry of AUTHOR, INSTALLATION, DATE-WRITTEN, DATE-COMPILED or SECURITY is any characters at all, written in Area B on one or more lines and never in Area A (Language Reference SC27-8713-03, p. 117), and a COPY or REPLACE in it, or where it can appear, is part of it (pp. 700, 708). So a comment-entry runs from its paragraph header's period to the next line, not a comment or blank line, with a character in Area A, whatever the lines between hold; NIST's OBNC1M tests this with a whole program written in Area B inside a SECURITY comment-entry",
1103 basis: Basis::Documented,
1104 oracle: Oracle::EnterpriseCobol,
1105 },
1106 Assumption {
1107 id: COMMENT_ENTRY_HEADERS,
1108 claim: "ironwork takes a line in an IDENTIFICATION DIVISION as a comment-entry paragraph's header when its first word is the paragraph's name and a period follows, in whatever column the name starts, though a paragraph header belongs in Area A (Language Reference SC27-8713-03, p. 55); with no period the paragraph's text is read as program text. A line with a hyphen in column 7 inside a comment-entry, which p. 117 does not permit, is taken as more of the comment-entry rather than refused: the manual does not say how severe Enterprise COBOL's message for either is",
1109 basis: Basis::Chosen,
1110 oracle: Oracle::EnterpriseCobol,
1111 },
1112 Assumption {
1113 id: COMMENT_ENTRY_REMARKS,
1114 claim: "Enterprise COBOL has no REMARKS paragraph: the Language Reference SC27-8713-03 names neither the paragraph (pp. 101, 117) nor the word anywhere. ironwork reads a REMARKS paragraph in an IDENTIFICATION DIVISION as it reads a comment-entry paragraph, as OS/VS COBOL did, where Enterprise COBOL presumably refuses the program",
1115 basis: Basis::Chosen,
1116 oracle: Oracle::EnterpriseCobol,
1117 },
1118 Assumption {
1119 id: CONTINUED_LITERAL_QUOTES,
1120 claim: "A continuation line of an alphanumeric or national literal left open at column 72 has a hyphen in column 7 and a quotation mark as its first nonblank character, and the literal resumes after that mark. When a literal's closing quotation mark is in column 72 and the continuation line starts with two, the pair stands for one quotation mark in one literal; otherwise, a quotation mark that starts a continuation line after a closed literal starts a second literal (Language Reference SC27-8713-03, p. 58). The rules hold for apostrophes alike, and ironwork gives the pair precedence where both could apply",
1121 basis: Basis::Documented,
1122 oracle: Oracle::EnterpriseCobol,
1123 },
1124 Assumption {
1125 id: COPY_SEARCH_ROUNDS,
1126 claim: "A COPY member is looked for in three rounds, each through every copy library in order (the program's own directory, then each -I directory) before the next begins: as NAME.cpy, .CPY, .copy and .COPY; then as .cbl, .CBL, .cob and .COB; then as the name alone; each extension with the name as written, then upper-cased, then lower-cased. In one z/OS UNIX directory IBM tries .cpy, .CPY, .cbl, .CBL, .cob and .COB, each with the name upper- and lower-cased (Language Reference SC27-8713-03, p. 705), and the name alone last (Programming Guide SC27-8714-03, p. 440), searching the current directory, the -I directories and SYSLIB's in turn (LR p. 705, PG p. 441); it does not say whether an earlier directory's .cbl comes before a later one's .cpy. A compile from JCL searches SYSLIB's data sets, then COPYLOC's (LR p. 705), which hold copybooks and not the library the program is read from, so the rounds take a copybook in any library before a program source. .copy is not IBM's, and the name as written, which IBM folds to upper case (PG p. 440), comes first, so that on a file system that ignores case a message names the member as the program spells it; the order differs from IBM's only where two files' names differ only in case",
1127 basis: Basis::Chosen,
1128 oracle: Oracle::EnterpriseCobol,
1129 },
1130 Assumption {
1131 id: COPY_NOT_THE_PROGRAM,
1132 claim: "The file being compiled is never a COPY member of its own compilation, at any depth of nesting: a compile from JCL reads the program from SYSIN and looks for members in SYSLIB and COPYLOC (Language Reference SC27-8713-03, p. 705), so a member named as the program comes from another file or is not found. The program's own directory is a copy library only as ironwork's stand-in for the current directory cob2 searches first (p. 705). Any other member a chain of COPY statements reaches while it is still being copied is refused, as a nested COPY cannot cause recursion (p. 697)",
1133 basis: Basis::Chosen,
1134 oracle: Oracle::EnterpriseCobol,
1135 },
1136 Assumption {
1137 id: COPY_LITERAL_AS_WRITTEN,
1138 claim: "A text-name written as a literal is looked for as written first, through every library, and then with extensions as a user-defined word is (COPY_SEARCH_ROUNDS). IBM takes a literal as the file name, relative path or absolute path it spells (Programming Guide SC27-8714-03, p. 440) and adds extensions only to a name that is not a literal (Language Reference SC27-8713-03, p. 705); the extensions are ironwork's, for sources written for compilers that add them",
1139 basis: Basis::Chosen,
1140 oracle: Oracle::EnterpriseCobol,
1141 },
1142 Assumption {
1143 id: COPY_DOUBLED_PERIOD,
1144 claim: "COPY X.. is refused, naming X.: a COPY statement ends with a separator period (Language Reference SC27-8713-03, p. 697), which is a period followed by a space (pp. 49-50), so the name is X., not X. A text-name or library-name for a data set holds only letters, digits and hyphens (p. 696); in z/OS UNIX directories any COBOL character may appear (p. 697), and X. would then name X..cpy and the like, not X's file. The manuals do not say whether Enterprise COBOL instead reads X and a separator period, leaving the second period in the text, as GnuCOBOL 3.2 does",
1145 basis: Basis::Chosen,
1146 oracle: Oracle::EnterpriseCobol,
1147 },
1148 Assumption {
1149 id: LINAGE_COUNTER,
1150 claim: "LINAGE IS n LINES [WITH FOOTING AT f] [LINES AT TOP t] [LINES AT BOTTOM b] describes a logical page of t + n + b lines, each page following the last with no spacing; TOP and BOTTOM default to 0, and each value is an unsigned integer, at most 99,999,999, or an unsigned integer data item (Language Reference SC27-8713-03, pp. 189-190, 747). OPEN OUTPUT or EXTEND takes all four for the first page, and a WRITE ... ADVANCING PAGE or a page overflow takes the data items' values again for the next page (p. 190). LINAGE-COUNTER, one for each LINAGE file and qualified by its file-name when there are two, has the PICTURE and USAGE of the page body's data item, or is binary with as many digits as its integer; OPEN sets it to 1, it is the line of the page body the printer is at, and no statement may change it (pp. 23-24, 70). A WRITE adds its ADVANCING lines to it, 1 without ADVANCING; a WRITE that would take it past the page body puts its line on the next page's first line, after the printer moves there (AFTER) or before (BEFORE), as ADVANCING PAGE does, and sets it to 1 (pp. 474-475). END-OF-PAGE, which needs LINAGE, runs once the line is written when LINAGE-COUNTER has reached the footing line or the page overflowed, only an overflow counting when FOOTING is not given, and NOT END-OF-PAGE otherwise, neither after a WRITE that failed (p. 475; Programming Guide SC27-8714-03, p. 178). LINAGE takes effect only for a file opened OUTPUT or EXTEND (p. 189), and an SD's does nothing (p. 190)",
1151 basis: Basis::Documented,
1152 oracle: Oracle::EnterpriseCobol,
1153 },
1154 Assumption {
1155 id: LINAGE_PAGE_MOVEMENT,
1156 claim: "A WRITE to a LINAGE file moves the paper in lines, never by a skip to a channel, the logical page not being the printer's form (Language Reference SC27-8713-03, p. 190). When OPEN ends the printer is at the first page's first line, t lines of top margin above line 1 of its page body, so the first WRITE moves those t lines besides its own; a WRITE that starts a new page, by overflow or ADVANCING PAGE, moves the lines left in the page body, the bottom margin, the next page's top margin and one line more, to that page's line 1. The lines are written as any WRITE ... ADVANCING writes them, as control characters and spacing records (PRINT_CONTROL_CHARACTER, PRINT_SPACING_RECORDS), so the margins are blank lines, and a text DD shows them as line feeds (TEXT_PRINT_LINES). OPEN and CLOSE write nothing: a file closed with no WRITE is empty, and the last page is not spaced out",
1157 basis: Basis::Chosen,
1158 oracle: Oracle::EnterpriseCobol,
1159 },
1160 Assumption {
1161 id: LINAGE_END_OF_PAGE,
1162 claim: "The end-of-page condition is judged by LINAGE-COUNTER once the WRITE is done, as the Language Reference words it (SC27-8713-03, p. 475): a WRITE ... ADVANCING PAGE, which leaves LINAGE-COUNTER at 1, raises it only when the next page's FOOTING is 1, whatever line it printed on, and a WRITE ... ADVANCING 0 LINES in the footing area raises it again",
1163 basis: Basis::Chosen,
1164 oracle: Oracle::EnterpriseCobol,
1165 },
1166 Assumption {
1167 id: LINAGE_VALUES,
1168 claim: "A page body under 1 line, a footing line outside the page body or a margin below 0 that data items give at OPEN or at a new page ends the run with ironwork's own abend naming the file: the Language Reference states the rule (SC27-8713-03, p. 189) but no file status for breaking it (Table 34, pp. 300-303), and the Language Environment Runtime Messages (SA38-0686-60) have no message for it. Integers that break it are compile errors",
1169 basis: Basis::Chosen,
1170 oracle: Oracle::EnterpriseCobol,
1171 },
1172 Assumption {
1173 id: LINAGE_EXTEND,
1174 claim: "OPEN EXTEND of a LINAGE file starts a new logical page as OPEN OUTPUT does, top margin first and LINAGE-COUNTER at 1 (Language Reference SC27-8713-03, p. 24), whatever line the data set's last page ended on",
1175 basis: Basis::Chosen,
1176 oracle: Oracle::EnterpriseCobol,
1177 },
1178 Assumption {
1179 id: LINAGE_COUNTER_BETWEEN_WRITES,
1180 claim: "LINAGE-COUNTER is zero until the file is first opened and keeps its value after CLOSE; a WRITE that fails leaves it and the page where they were; OPEN INPUT and I-O set it to 1 as OPEN OUTPUT does, since the Language Reference says so of any OPEN (SC27-8713-03, p. 24)",
1181 basis: Basis::Chosen,
1182 oracle: Oracle::EnterpriseCobol,
1183 },
1184 Assumption {
1185 id: PRINT_FILE_UPDATE,
1186 claim: "A print file opened I-O under ADV holds the control character a byte before each FD record, as it was written (PRINT_CONTROL_CHARACTER): READ skips the byte as PRINT_CONTROL_RUN_TIME says, REWRITE writes the record behind the byte it was read with, REWRITE having no ADVANCING phrase, and WRITE, with or without ADVANCING, fails with file status 48, a sequential file's WRITE needing OUTPUT or EXTEND (Language Reference SC27-8713-03, pp. 471, 476; Table 34, p. 302); LINAGE has no effect on it (p. 189). IBM does not say what REWRITE does with the byte",
1187 basis: Basis::Chosen,
1188 oracle: Oracle::EnterpriseCobol,
1189 },
1190 Assumption {
1191 id: ENTRY_IN_SEQUENCE,
1192 claim: "Control that reaches an ENTRY statement in the program's own sequence passes it as it passes CONTINUE: nothing is bound and no storage changes. The Language Reference says only where a CALL of the entry begins, at the first executable statement after it (SC27-8713-03, pp. 339-340)",
1193 basis: Basis::Chosen,
1194 oracle: Oracle::EnterpriseCobol,
1195 },
1196 Assumption {
1197 id: ENTRY_CALLS,
1198 claim: "A CALL of an ENTRY name begins at the statement after the ENTRY and binds the entry's USING list alone, so a LINKAGE item only the PROCEDURE DIVISION USING names has no address (Language Reference SC27-8713-03, pp. 264, 320, 339-340). A static CALL, of a literal under NODYNAM, enters the one copy of the program that its PROGRAM-ID enters, in its last-used state (Programming Guide SC27-8714-03, pp. 548, 553-554); a dynamic CALL, of an identifier or of a literal under DYNAM, gets for each entry name a copy of the program with WORKING-STORAGE of its own, as if a separate compile unit were called (p. 560), which a CANCEL of that name resets, where p. 549 says only that a second entry point must not be called dynamically without a CANCEL between. CALL finds an entry name among the programs of the source and those already loaded, then in a program library as a member of that name, an alias, as NAME(ALIAS) or binder ALIAS statements make one (pp. 548, 560). Under --dialect gnucobol every entry name enters the program's one copy, as cobc's do: a dynamic CALL of an entry name shares the program's WORKING-STORAGE, and a CANCEL of an entry name does nothing",
1199 basis: Basis::Chosen,
1200 oracle: Oracle::EnterpriseCobol,
1201 },
1202 Assumption {
1203 id: ALTERED_GO_TO_RESET,
1204 claim: "Altered GO TOs are put back as written whenever the program's WORKING-STORAGE is initialized: its first CALL, the first after a CANCEL of it, and every CALL of an INITIAL program (Language Reference SC27-8713-03, pp. 103, 318), and not on another CALL (Programming Guide SC27-8714-03, p. 548). Those of an independent segment, priority 50 to 99, are put back when control reaches the segment from a paragraph of another priority by falling into it, GO TO, PERFORM or a SORT or MERGE procedure, but not when a PERFORM made from it returns, nor when a CALL of the program begins in it: p. 265 says 'from a segment with a different priority-number', p. 318 'from another independent segment', and neither names a PERFORM's return",
1205 basis: Basis::Chosen,
1206 oracle: Oracle::EnterpriseCobol,
1207 },
1208 Assumption {
1209 id: DISPLAY_STREAM,
1210 claim: "DISPLAY writes to standard output as a program under z/OS UNIX does with its OUTDD ddname unallocated and _IGZ_SYSOUT unset: one stream of characters, a newline after each DISPLAY except one WITH NO ADVANCING (Programming Guide SC27-8714-03, pp. 36-37; Language Reference SC27-8713-03, pp. 333-335), and UPON any device writes to the same stream. To a ddname IBM writes a record for each DISPLAY, whose first byte is ' ', or '+' after one WITH NO ADVANCING (Programming Guide p. 37); that route is not modelled",
1211 basis: Basis::Chosen,
1212 oracle: Oracle::EnterpriseCobol,
1213 },
1214 Assumption {
1215 id: RANDOM_GENERATOR,
1216 claim: "FUNCTION RANDOM is Park and Miller's minimal standard generator. An argument n starts a sequence at state n mod 2147483646 + 1, a first reference without one starts it as 0 does, and each reference sets the state s to 16807 s mod 2147483647 and returns the new state divided by 2147483647 in long HFP, which is exclusively between zero and one; so the arguments 0 to 2,147,483,645 give distinct sequences and larger ones repeat them. IBM documents the interface (Language Reference SC27-8713-03, p. 629), a long floating-point result under either ARITH (Programming Guide SC27-8714-03, p. 58), and that its generator is not CEERAN0's, 950706376 s mod 2147483647 (Programming Reference SA38-0683-60, pp. 343-344), but not the generator. A fractional argument is truncated to an integer, and a negative one ends the run with abend IRONWORK",
1217 basis: Basis::Chosen,
1218 oracle: Oracle::EnterpriseCobol,
1219 },
1220 Assumption {
1221 id: ZERO_DIVISOR_CHECK,
1222 claim: "A zero divisor that no ON SIZE ERROR phrase takes, in an arithmetic statement without one or in an expression outside an arithmetic statement (a condition, a subscript, a reference modifier), is the program check of the instruction the compiler divides with: HFP divide, S0CF, when the expression is evaluated in floating point; fixed-point divide, S0C9, when the dividend and the divisor are made only of integer binary items and integer literals, at least one an item; decimal divide, S0CB, otherwise. The size error condition belongs to the arithmetic statements alone, and with ON SIZE ERROR any zero divisor, floating-point too, is one (Language Reference SC27-8713-03, p. 296; Programming Guide SC27-8714-03, p. 242); the manuals do not say which instructions the compiler divides with",
1223 basis: Basis::Chosen,
1224 oracle: Oracle::EnterpriseCobol,
1225 },
1226 Assumption {
1227 id: ERROR_DECLARATIVE_MODE,
1228 claim: "An EXCEPTION/ERROR procedure for an open mode serves a file open in that mode, and for OPEN one being opened in it, an OPEN of a file already open included (Language Reference SC27-8713-03, pp. 417, 714). A file that is not open, as for a READ, WRITE or CLOSE before its OPEN, is in no mode, so only a procedure that names it serves it. A procedure that names the file comes first (p. 714), and two procedures for one file, or for one open mode, are refused, as p. 714 forbids simultaneous requests for two",
1229 basis: Basis::Chosen,
1230 oracle: Oracle::EnterpriseCobol,
1231 },
1232 Assumption {
1233 id: ERROR_DECLARATIVE_STATUSES,
1234 claim: "Every I/O status whose first digit is not 0 runs the file's EXCEPTION/ERROR procedure once FILE STATUS holds it, unless the statement's AT END or INVALID KEY phrase takes a 1x or 2x status, and then no procedure runs (Language Reference SC27-8713-03, pp. 299, 303-304, 432, 714); a 0x status runs none. The procedure returns control to the end of the statement, and NOT AT END and NOT INVALID KEY are not run: none of ironwork's statuses is a critical error, after which p. 714 says control does not return. The implicit CLOSE at the end of the run or at CANCEL runs no procedure (Programming Guide SC27-8714-03, pp. 179, 204, 219)",
1235 basis: Basis::Chosen,
1236 oracle: Oracle::EnterpriseCobol,
1237 },
1238 Assumption {
1239 id: SORT_FILE_DECLARATIVE,
1240 claim: "An EXCEPTION/ERROR procedure serves the OPEN, READ, WRITE and CLOSE a SORT or MERGE does of its USING and GIVING files (Language Reference SC27-8713-03, pp. 457-458); the end of a USING file runs none. After it, that file's processing ends: a USING file gives the records read before the failure and is closed, a GIVING file is closed, and the operation goes on with SORT-RETURN 0, unless the procedure moved 16 to SORT-RETURN, which stops it at once with SORT-RETURN 16, as the Programming Guide has the procedure do to report the failure (SC27-8714-03, pp. 232, 234-235). A file with no procedure fails as SORT_FILE_FAILURE says. Under FASTSRT a USING or GIVING file that an INPUT, OUTPUT or file-specific procedure serves is COBOL's (p. 233), which FASTSRT_FILES does not list",
1241 basis: Basis::Chosen,
1242 oracle: Oracle::EnterpriseCobol,
1243 },
1244 Assumption {
1245 id: DEBUG_RUNTIME_OPTION,
1246 claim: "USE FOR DEBUGGING procedures run only in a program compiled WITH DEBUGGING MODE and run under the Language Environment runtime option DEBUG, which -debug stands for; NODEBUG, the default, keeps them from running, and debugging lines, once compiled, run under either (Language Reference SC27-8713-03, pp. 771-772; Programming Guide SC27-8714-03, pp. 431, 446). Without WITH DEBUGGING MODE both are comments, and a contained program has the mode of the program containing it (LR pp. 121, 772)",
1247 basis: Basis::Documented,
1248 oracle: Oracle::EnterpriseCobol,
1249 },
1250 Assumption {
1251 id: DEBUG_LINE_NUMBER,
1252 claim: "DEBUG-LINE holds, in six digits with leading zeros, the number of the line the statement starts on in its own source file. Under NONUMBER, the default, IBM puts the compiler-generated number there (Language Reference SC27-8713-03, p. 19), the listing's line number, which counts the lines of COPY members too: a statement after a COPY, or in a member, is numbered otherwise by IBM",
1253 basis: Basis::Chosen,
1254 oracle: Oracle::EnterpriseCobol,
1255 },
1256 Assumption {
1257 id: DEBUG_LINE_STATEMENT,
1258 claim: "The statement DEBUG-LINE names is the one that sent control to the procedure (Language Reference SC27-8713-03, p. 20): the PERFORM on each repetition, the GO TO, the SORT or MERGE, the input-output statement whose condition ran a USE procedure; for fall through, the statement last started in the procedure before, or the section header control passed through. CONTINUE, EXIT, NEXT SENTENCE and a separator period carry no position in ironwork, so after one of them DEBUG-LINE names the statement before it",
1259 basis: Basis::Chosen,
1260 oracle: Oracle::EnterpriseCobol,
1261 },
1262 Assumption {
1263 id: DEBUG_CONTENTS_LENGTH,
1264 claim: "DEBUG-CONTENTS is 30 characters, so DEBUG-ITEM is 86 bytes. The Language Reference gives it as PICTURE X(n) (SC27-8713-03, p. 19), and the procedures Enterprise COBOL debugs put at most 13 characters in it (p. 20)",
1265 basis: Basis::Chosen,
1266 oracle: Oracle::EnterpriseCobol,
1267 },
1268 Assumption {
1269 id: DEBUG_NAME_FORM,
1270 claim: "DEBUG-NAME is the procedure-name as the USE FOR DEBUGGING sentence writes it, a section that qualifies it joined by OF (Language Reference SC27-8713-03, p. 19); under ALL PROCEDURES it is the procedure's own name, unqualified. Control entering a section runs the section's debugging procedure, then, by fall through, that of its first paragraph",
1271 basis: Basis::Chosen,
1272 oracle: Oracle::EnterpriseCobol,
1273 },
1274 Assumption {
1275 id: DEBUGGING_SECTION_REFERENCES,
1276 claim: "A debugging section may PERFORM or GO TO a procedure of another debugging section, and of an EXCEPTION/ERROR section: p. 771 of the Language Reference (SC27-8713-03) forbids referring to a procedure in a debugging section from a statement outside of the debugging section, read here as outside every debugging section, as the CCVS85 DB tests assume, and p. 716 forbids references to nondeclarative procedures only. Without WITH DEBUGGING MODE a debugging section is a comment in full, header and USE sentence included, so its names are not defined and its text is not checked (p. 772)",
1277 basis: Basis::Chosen,
1278 oracle: Oracle::EnterpriseCobol,
1279 },
1280 Assumption {
1281 id: GLOBAL_DECLARATIVES,
1282 claim: "A file statement of a contained program that has no EXCEPTION/ERROR procedure of its own for the file, by name or by open mode, runs the first USE GLOBAL procedure of the programs containing it, innermost out, for the file and then for the mode (Language Reference SC27-8713-03, pp. 714-715), as a procedure of the program that declares it: over that program's storage as it stood when control left it, its LINKAGE addresses included, with PERFORMs of its own. Control comes back after the statement; STOP RUN in the procedure ends the run, and GO TO out of it, GOBACK and EXIT PROGRAM (which p. 714 forbids while a declarative of a nested program is active) are refused when reached. USE GLOBAL BEFORE REPORTING for a report group of a contained program without its own procedure for it would serve another program's report (Report Writer Precompiler SC26-4301-04, 4.7.2 rule 5 and 4.7.3 rule 4) and is refused in a program that contains others",
1283 basis: Basis::Chosen,
1284 oracle: Oracle::EnterpriseCobol,
1285 },
1286 Assumption {
1287 id: SYNC_SUBORDINATE_GROUP,
1288 claim: "SYNCHRONIZED on a group at level 02 to 49 synchronizes each elementary item within it, as the clause does on a level-01 group; the Language Reference allows it on elementary items and level-01 groups (SC27-8713-03, p. 231) and says nothing of other groups",
1289 basis: Basis::Chosen,
1290 oracle: Oracle::EnterpriseCobol,
1291 },
1292 Assumption {
1293 id: SYNC_ONLY_WHEN_WRITTEN,
1294 claim: "A binary, COMP-1, COMP-2, POINTER or INDEX item is aligned only when SYNCHRONIZED applies to it. Table 15 of the Language Reference says a subordinate binary item is aligned on 2 or 4 bytes 'when the synchronized clause is not specified', and in the same row that 'when SYNCHRONIZED is not specified for binary items, no space is reserved for slack bytes' (SC27-8713-03, p. 232); ironwork takes the second, and the first as a slip for 'specified'. A synchronized binary item of 10 to 18 digits is aligned on 4 bytes, as the slack-byte algorithm gives (p. 233), not on 8",
1295 basis: Basis::Chosen,
1296 oracle: Oracle::EnterpriseCobol,
1297 },
1298 Assumption {
1299 id: SYNC_SLACK_OWNER,
1300 claim: "Slack bytes before a synchronized item belong to the group of the elementary item before it (Language Reference SC27-8713-03, p. 234), so a group that ended just before them grows by them; but when that group is a table or a redefinition, or ends earlier than the slack begins, the slack bytes stay in the group that holds it and the table's occurrences keep their length",
1301 basis: Basis::Chosen,
1302 oracle: Oracle::EnterpriseCobol,
1303 },
1304 Assumption {
1305 id: SYNC_REDEFINES_REFUSED,
1306 claim: "A synchronized item that starts a redefinition where its boundary would need slack bytes stops the compile. The Language Reference says such an item must not need them and that the redefined item must be aligned for it (SC27-8713-03, pp. 232-233), but not what the compiler does when a program breaks the rule",
1307 basis: Basis::Chosen,
1308 oracle: Oracle::EnterpriseCobol,
1309 },
1310 Assumption {
1311 id: DECIMAL_COMMA_SEPARATOR,
1312 claim: "Under DECIMAL-POINT IS COMMA a comma between digits, or after a space, parenthesis or sign and before a digit, is a numeric literal's decimal point, and a comma after any other word is a separator even with a digit after it, so T(1,2) has the one subscript 1,2 and T(I,2) has two. The Language Reference says only that a separator comma is a comma followed by a space (SC27-8713-03, p. 49) and that the clause exchanges the comma's and the period's functions in numeric literals (p. 131); a period before a digit is then an error",
1313 basis: Basis::Chosen,
1314 oracle: Oracle::EnterpriseCobol,
1315 },
1316 Assumption {
1317 id: DECIMAL_COMMA_DISPLAY_LITERAL,
1318 claim: "DISPLAY of a numeric literal writes it as the program wrote it, so under DECIMAL-POINT IS COMMA its decimal point is a comma; the Language Reference says nothing of how DISPLAY shows a numeric literal (SC27-8713-03, pp. 333-334). Under --dialect gnucobol it writes the literal without its decimal point, as cobc does: 1.5 shows as 15 and -0.25 as -025",
1319 basis: Basis::Chosen,
1320 oracle: Oracle::EnterpriseCobol,
1321 },
1322 Assumption {
1323 id: ARITH_DIGIT_LIMITS,
1324 claim: "A numeric or numeric-edited PICTURE whose digit positions, scaling positions P included, exceed 18 under ARITH(COMPAT) or 31 under ARITH(EXTEND), and a numeric literal with more digits than that, stop the compile (Language Reference SC27-8713-03, pp. 45, 209, 217-218; Programming Guide SC27-8714-03, p. 349). Neither manual gives the diagnostic's severity, and P is counted for every numeric item, where the Language Reference counts it for numeric-edited items and arithmetic operands (p. 209)",
1325 basis: Basis::Chosen,
1326 oracle: Oracle::EnterpriseCobol,
1327 },
1328 Assumption {
1329 id: MULTIPLE_RESULTS,
1330 claim: "An arithmetic statement with several receivers computes what they share once, before any is stored, and each receiver in turn then takes it or combines it with its own current value, its subscripts evaluated then (Language Reference SC27-8713-03, p. 298). For ADD, SUBTRACT, MULTIPLY and DIVIDE without GIVING the shared part is the operands other than the receiver; for COMPUTE and the GIVING forms it is the whole expression, so a receiver that COMPUTE names twice gets the same result twice",
1331 basis: Basis::Chosen,
1332 oracle: Oracle::EnterpriseCobol,
1333 },
1334 Assumption {
1335 id: ALTER_DEBUGGING,
1336 claim: "Once an ALTER has run, the debugging section that serves each paragraph it alters runs, once for each TO [PROCEED TO] phrase in the order written, with DEBUG-LINE the ALTER, DEBUG-NAME the altered paragraph as DEBUG_NAME_FORM gives it, and DEBUG-CONTENTS the procedure-name after TO PROCEED TO, a qualifier after OF (Language Reference SC27-8713-03, pp. 19-20, 716). A procedure named only after TO PROCEED TO gets no debugging section from the ALTER: p. 716 says an ALTER 'referring to the named procedure', and Table 2 on p. 20 has an ALTER row for procedure-name-1 alone, as CCVS85 DB105A expects under ALL PROCEDURES. Under ALL PROCEDURES an ALTER in the declaratives runs none, as p. 716 says; there an ALTER of a paragraph that a USE FOR DEBUGGING names still runs its section, as p. 716 makes no exception for it",
1337 basis: Basis::Chosen,
1338 oracle: Oracle::EnterpriseCobol,
1339 },
1340 Assumption {
1341 id: PERFORM_RETURN_POINTS,
1342 claim: "An out-of-line PERFORM arms a return point at the end of its range's last paragraph (Language Reference SC27-8713-03, p. 419), one per activation, since a CALL resets return points (Programming Guide SC27-8714-03, p. 547). Control that passes that end by any path, falling through or by GO TO, returns to the PERFORM, so PERFORM B THRU A with A before B returns when control reaches the end of A, and a range that passes the end of another active PERFORM's range returns there to that PERFORM. Neither manual says what a PERFORM that control leaves by GO TO leaves behind: its return point stays armed, as the Programming Guide's warning against ranges that keep control from the end implies (p. 772), until control passes it and returns after that PERFORM, which then puts back the point it displaced; ironwork refuses at run time to return so into a PERFORM that repeats or is inside another statement. EXIT SECTION goes to the end of the section, past the return point of a performed paragraph in it (LR p. 345)",
1343 basis: Basis::Chosen,
1344 oracle: Oracle::EnterpriseCobol,
1345 },
1346 Assumption {
1347 id: FLOAT_FUNCTION_ARGUMENTS,
1348 claim: "A floating-point argument, item or expression, is allowed wherever a function takes a numeric argument and refused where it takes an integer (Language Reference SC27-8713-03, p. 507). INTEGER and INTEGER-PART of one return an integer of 30 digits, 31 under ARITH(EXTEND), and ABS, MAX, MIN, RANGE, REM and SUM with one are evaluated in floating point and return it (Programming Guide SC27-8714-03, pp. 799 and 801). The guide names REM a mixed function where the Language Reference types it numeric (p. 633), and gives the precision only of floating-point functions: a mixed function is evaluated here in long floating point, extended under ARITH(EXTEND), as they are",
1349 basis: Basis::Chosen,
1350 oracle: Oracle::EnterpriseCobol,
1351 },
1352 Assumption {
1353 id: FLOAT_FUNCTION_ROUNDING,
1354 claim: "The floating-point intrinsic functions are computed in binary to 128 bits and rounded to the nearest long HFP value under ARITH(COMPAT), extended under ARITH(EXTEND), ties away from zero; a fixed-point argument is first converted to HFP of that precision (C5). IBM computes SQRT, EXP, EXP10, LOG, LOG10 and the trigonometric functions with Language Environment's math services, CEESDSQT and the rest (Programming Guide SC27-8714-03, p. 58), whose results can differ from the nearest value in the last hexadecimal digit. ANNUITY, PRESENT-VALUE and the statistics functions, which have no such service, are computed the same way, where IBM's generated code may truncate at each HFP step",
1355 basis: Basis::Chosen,
1356 oracle: Oracle::EnterpriseCobol,
1357 },
1358 Assumption {
1359 id: FLOATING_POINT_FUNCTIONS,
1360 claim: "ACOS, ANNUITY, ASIN, ATAN, COS, LOG, LOG10, MEAN, MEDIAN, MIDRANGE, PRESENT-VALUE, RANDOM, SIN, SQRT, STANDARD-DEVIATION, TAN and VARIANCE are floating-point functions, as earlier Programming Guides listed them; E, PI, EXP, EXP10 and NUMVAL-F are, as the Language Reference says (SC27-8713-03, pp. 553-557, 609); SECONDS-FROM-FORMATTED-TIME is, as its example's inexact result shows (p. 507), and SECONDS-PAST-MIDNIGHT with it. ABS, MAX, MIN, RANGE, REM and SUM are floating point when any argument is (Programming Guide, p. 799; C100). An expression holding a floating-point function is evaluated in floating point (pp. 62-63). NUMVAL and NUMVAL-C are too, long under ARITH(COMPAT) and extended under ARITH(EXTEND), as the Programming Guide says (SC27-8714-03, p. 115); COMBINED-DATETIME is long whatever ARITH says (p. 541)",
1361 basis: Basis::Recalled,
1362 oracle: Oracle::EnterpriseCobol,
1363 },
1364 Assumption {
1365 id: FUNCTION_DOMAIN,
1366 claim: "An argument outside a function's domain ends the run with abend IRONWORK: SQRT of a negative number, LOG or LOG10 of zero or less, ASIN or ACOS beyond -1 to +1, ANNUITY with a negative rate or periods that are not a positive integer, PRESENT-VALUE at a rate of -1 or less, FACTORIAL beyond 28 (29 under ARITH(EXTEND)), a century window whose end year is outside 1700 to 9999, HEX-TO-CHAR or BIT-TO-CHAR of other characters or of a length that is not a multiple of 2 or 8. IBM leaves such values undefined (Language Reference SC27-8713-03, p. 500) and Language Environment's math services signal a condition. SIN, COS and TAN of an argument beyond 2^63 times pi/2, which ironwork does not reduce, end the run the same way. A result beyond HFP's range is an exponent overflow, S0CC, and one below it zero (C8)",
1367 basis: Basis::Chosen,
1368 oracle: Oracle::EnterpriseCobol,
1369 },
1370 Assumption {
1371 id: NUMVAL_TEST_RULES,
1372 claim: "TEST-NUMVAL, TEST-NUMVAL-C and TEST-NUMVAL-F follow the formats of NUMVAL, NUMVAL-C and NUMVAL-F (Language Reference SC27-8713-03, pp. 605-609, 651-655), choosing where they are silent: CR and DB in either case; NUMVAL-C's grouping separator only between digits and before the decimal point; NUMVAL-F's E in either case, spaces allowed around it, its exponent sign optional, and the 16-digit mantissa limit with an exponent not checked; a string that stops short, or holds only spaces, gives its length + 1. NUMVAL-C's currency string defaults to $. NUMVAL-F of a string that breaks them returns zero, as NUMVAL does here",
1373 basis: Basis::Chosen,
1374 oracle: Oracle::EnterpriseCobol,
1375 },
1376 Assumption {
1377 id: FUNCTION_CLOCK,
1378 claim: "SECONDS-PAST-MIDNIGHT, FORMATTED-CURRENT-DATE, whose offset is therefore +0000, and the year of execution that YEAR-TO-YYYY, DATE-TO-YYYYMMDD and DAY-TO-YYYYDDD window by, read the run unit's clock as UTC to the hundredth of a second, as CURRENT-DATE does here (+0000); z/OS gives local time, and finer seconds (Language Reference SC27-8713-03, p. 631)",
1379 basis: Basis::Chosen,
1380 oracle: Oracle::EnterpriseCobol,
1381 },
1382 Assumption {
1383 id: UUID4_SOURCE,
1384 claim: "UUID4 takes its 122 random bits from the process's randomly keyed hasher over the clock, not from a cryptographic generator; IBM uses the Message-Security-Assist random number facility where the machine has it (Language Reference SC27-8713-03, p. 669). The version and variant bits are set and the string is lowercase, as IBM's example shows",
1385 basis: Basis::Chosen,
1386 oracle: Oracle::EnterpriseCobol,
1387 },
1388 Assumption {
1389 id: FORMATTED_DATETIME_RULES,
1390 claim: "The FORMATTED functions, INTEGER-OF-FORMATTED-DATE, SECONDS-FROM-FORMATTED-TIME and TEST-FORMATTED-DATETIME take the formats of the Language Reference (SC27-8713-03, pp. 504-506), choosing where it is silent or inconsistent: the decimal separator of a fractional-seconds format appears in the data, as its rules say, though several of its examples omit it (pp. 561, 566, 568, 629); fractional seconds are truncated; a comma may stand for the period; a UTC format moves the date as well as the time by the offset; a week may be 53 where the ISO year has 53 weeks, though p. 506 says 01 to 52; an offset sign of 0 takes only 00 hours and minutes; INTEGER-OF-FORMATTED-DATE reads the date part alone, as p. 579 says the time part does not change its result. TEST-FORMATTED-DATETIME names the first position at which a field can no longer be in range, or a value longer than its format errs at the first extra character. A format that is not one of IBM's, which Enterprise COBOL refuses at compile time, ends the run here",
1391 basis: Basis::Chosen,
1392 oracle: Oracle::EnterpriseCobol,
1393 },
1394 Assumption {
1395 id: ROUNDED_EXTRA_PLACE,
1396 claim: "A receiver named with ROUNDED counts in dmax with one decimal place more than it holds, so a quotient, or an intermediate cut back to dmax places, keeps the digit that rounding reads: DIVIDE 44.1 INTO a PIC 9(4)V9 of 1661.7 ROUNDED gives 37.7, as CCVS85 NC117A and NC171A expect. The Programming Guide says only that under ROUNDED one more decimal place, and one more integer place, might be carried for accuracy if necessary (SC27-8714-03, p. 794); the Language Reference's ROUNDED phrase compares the result's fraction with the receiver's (SC27-8713-03, p. 296). Under --dialect gnucobol the extra place counts in the statement's last operation alone, whose result the receivers take, and every operation below it carries dmax with each receiver's own places, as cobc -std=ibm truncates intermediate results to dmax and computes the last one exactly: COMPUTE D ROUNDED = D + E / 3 keeps E / 3 to D's two places",
1397 basis: Basis::Chosen,
1398 oracle: Oracle::EnterpriseCobol,
1399 },
1400 Assumption {
1401 id: CURRENCY_SIGNS,
1402 claim: "Once a program has a CURRENCY SIGN clause, $ is a currency symbol in its PICTUREs only if a clause names it: the Language Reference says the currency symbol is $ or the character a clause or the CURRENCY option specifies, and that the clause overrides the option (SC27-8713-03, pp. 130, 212), not that $ stays. A floating currency string of a value longer than one character ends in the position left of the first digit shown, the first currency position holding the whole value (p. 210). NUMVAL-C and TEST-NUMVAL-C without argument-2 take as cs the value of the program's only CURRENCY SIGN clause, where p. 616 names the currency symbol, the CURRENCY option's character when there is no clause (p. 212; C211), and $ otherwise. A hexadecimal currency sign literal is read as C141 says",
1403 basis: Basis::Chosen,
1404 oracle: Oracle::EnterpriseCobol,
1405 },
1406 Assumption {
1407 id: JSON_GENERATE_RULES,
1408 claim: "JSON GENERATE (Language Reference SC27-8713-03, pp. 369-382) writes zoned, packed, binary, index and internal floating-point items as JSON numbers and every other elementary item as a string, choosing where the manual is silent: a table element that a SUPPRESS ... WHEN phrase leaves out is left out of its array, and a table all of whose elements are left out is left out; the members of an unnamed group join its parent's object; a COMP-1 or COMP-2 value takes the digits of its exact HFP value rounded to 9 or 18 significant digits; a character an EBCDIC ENCODING cannot hold becomes X'3F'; when the receiver is too small it holds the leading bytes of the document, whole characters for a national receiver, and COUNT names that many character positions; JSON-CODE and JSON-STATUS are declared in each program that has the statement rather than as GLOBAL in the outermost one",
1409 basis: Basis::Chosen,
1410 oracle: Oracle::EnterpriseCobol,
1411 },
1412 Assumption {
1413 id: XML_PARSE_RULES,
1414 claim: "XML PARSE runs as under XMLPARSE(XMLSS); XMLPARSE(COMPAT) and VALIDATING are not supported. Where the manuals are silent: a document that is not well formed gives XML-CODE with z/OS XML System Services' return code 12 and the non-validating parser's reason (SA38-0681-50, Appendix B): 2004 when it ends before the root's end tag, 2019 with no root, 3000 a duplicate attribute, 3008 -- in a comment, 3022 < in an attribute value, 3028 a bad character reference, 3035 a mismatched end tag, 3060 a malformed XML declaration or a later processing instruction named xml, 3061 an undeclared entity, 3062 any other character out of place, text after the root included, 3065 a second root; an undeclared prefix is Enterprise COBOL's warning 00040800 or 00040801, reported before the name's own event, and the parse goes on only when the processing procedure sets XML-CODE to zero, as the Programming Guide's Table 83 shows (pp. 656-658); after any other exception XML-CODE keeps the parser's code whatever the procedure sets. At END-OF-INPUT, XML-CODE 1 takes identifier-1's content, evaluated again, as the next segment, and any other value ends the input, so an unfinished document is then an exception. Markup a segment ends inside is held until it is complete, while content, comments and processing-instruction data are reported in parts, the target again before each later part (Programming Guide SC27-8714-03, pp. 652-653); START-OF-CDATA-SECTION waits for a character after <![CDATA[; namespace declarations are reported after START-OF-ELEMENT and before the attributes; a character reference to a character the document's code page lacks is a NATIONAL-CHARACTER event; XML-TEXT for EXCEPTION holds the document up to the error; the registers' fragments live in run-unit storage released after each event",
1415 basis: Basis::Chosen,
1416 oracle: Oracle::EnterpriseCobol,
1417 },
1418 Assumption {
1419 id: XML_GENERATE_RULES,
1420 claim: "Where the Language Reference (SC27-8713-03, pp. 484-494) and the Programming Guide (SC27-8714-03, pp. 663-669, 817) are silent, XML GENERATE: trims alphanumeric-edited and numeric-edited values of trailing spaces only, as items of class alphanumeric; puts an unnamed group's members in its parent's element; places TYPE CONTENT items in the parent's content in the order of the data description, among its child elements; writes an element with no content as a start and an end tag, never an empty-element tag, as the Programming Guide's examples show, and keeps a group with no attributes and no content unless a SUPPRESS phrase is given; lets an item's own SUPPRESS ... WHEN decide for it in place of every EVERY phrase; writes a value holding a character XML 1.0 cannot hold as the item's storage in upper-case hexadecimal under its name prefixed hex., sets 417 and goes to ON EXCEPTION once the whole document is written, with 400 before 417 and 417 before 418; drops a namespace's trailing spaces, escapes it as an attribute value, gives 416 for a character XML cannot hold, and ignores NAMESPACE-PREFIX when the namespace is empty; names an EBCDIC CCSID in the XML declaration as IBM- and at least three digits, as the Programming Guide's IBM-037 shows; makes a national item in a document in an EBCDIC code page exception 420 at run time, where the manual makes it a compile-time rule, and a character the code page lacks its ? with 418; and leaves the receiver and COUNT unchanged for 411, 414, 415, 416, 419 and 420",
1421 basis: Basis::Chosen,
1422 oracle: Oracle::EnterpriseCobol,
1423 },
1424 Assumption {
1425 id: JSON_PARSE_RULES,
1426 claim: "Where the Language Reference (SC27-8713-03, pp. 382-396) and the Programming Guide (SC27-8714-03, pp. 609-617, 819-822) are silent, JSON PARSE: takes only an object or an array as the outermost value (100 otherwise); gives 104 for an object or array where an elementary item stands, and for any other value where a group or table stands; lets a pair that names a suppressed item pass without status 2; reads an unnamed group's members as its parent's and leaves an unnamed table alone; compares duplicate pairs as parsed values, 4 when equal and 103 when not, the first staying; ends the walk at an exception, leaving what it set, with JSON-STATUS as far as it got, and gives 106 when no value reached an elementary item or a null; sets an INDICATING indicator whenever its item's pair is met, the first value for null and the second otherwise; reads a string for a numeric receiver as spaces, a sign, digits with at most one decimal point, and spaces (the form of APAR PH65883); moves a number into an alphanumeric or national receiver only as an integer, as MOVE moves an integer literal, the sign dropped; truncates fraction digits beyond the receiver's; sets 128 when a numeric receiver loses high-order digits, 256 when a string loses characters other than spaces or an integer loses digits, and 512 with X'3F' for each character the code page lacks; rounds a number once into COMP-1 or COMP-2; accepts WITH DETAIL without issuing the IGZ messages; and, as JSON GENERATE does, takes a table named without its last subscript as the whole table",
1427 basis: Basis::Chosen,
1428 oracle: Oracle::EnterpriseCobol,
1429 },
1430 Assumption {
1431 id: DLI_TRANSLATION,
1432 claim: "EXEC DLI is read as the IMS translator reads it (IMS Application Programming: EXEC DLI Commands for CICS and IMS, SC18-7811-04, chapters 4-6): a command by its name or longer spelling (DELETE, from the book's C sample, for DLET), its options checked against the book's Format diagrams (pp. 35-81), GHU, GHN and GHNP taking their Get command's (p. 102), AIB allowed wherever PCB is (p. 5), and GMSG, ICMD and RCMD, which the book leaves to the Operations Guide, not checked; the data an option names and the right-hand side of each WHERE comparison declared; SEGMENT or PSB in double parentheses naming an area. The book names no relational operators or connectors beyond its examples' =, >=, >, < and AND and OR, so EQ, NE, GT, GE, LT, LE, <=, ¬= and the symbols &, |, * and + are accepted too. The DL/I interface block is declared at the head of WORKING-STORAGE with the book's COBOL labels over the 40 bytes its C declaration gives (p. 6), the unnamed bytes FILLER, unless the program declares DIBSTAT itself; the book does not name the 01 level, so it is DLZDIB, DL/I's name for the block, and the translator's IS GLOBAL is left out. A command is checked, not run: reaching one ends the run",
1433 basis: Basis::Chosen,
1434 oracle: Oracle::EnterpriseCobol,
1435 },
1436 Assumption {
1437 id: PASSWORD_IGNORED,
1438 claim: "A SELECT's PASSWORD clause is read and has no effect: the files ironwork for COBOL opens carry no VSAM passwords, so the password items are neither checked against the file nor required to hold one before OPEN (Language Reference SC27-8713-03, p. 152, where IBM requires a valid password for a VSAM file)",
1439 basis: Basis::Chosen,
1440 oracle: Oracle::EnterpriseCobol,
1441 },
1442 Assumption {
1443 id: CORRESPONDING_PAIRS,
1444 claim: "MOVE, ADD and SUBTRACT CORRESPONDING name one sending and one receiving group, neither reference-modified nor a level-66, 77 or 88 item. Two items under them correspond when they have the same name and the same qualifiers up to the groups, neither is FILLER, and neither is described with RENAMES, REDEFINES, OCCURS or a USAGE of INDEX, POINTER, FUNCTION-POINTER, PROCEDURE-POINTER or OBJECT REFERENCE, which also leaves out everything under such an item; for MOVE at least one is elementary and the move is valid in IBM's table of elementary moves, and for ADD and SUBTRACT both are elementary numeric. Each MOVE pair gives the result of its own MOVE, and ROUNDED and the SIZE ERROR phrases apply to every ADD or SUBTRACT pair, ON SIZE ERROR running once after all of them (Language Reference for Enterprise COBOL 6.4, 'CORRESPONDING phrase', 'MOVE statement', 'ADD statement', 'SUBTRACT statement', 'SIZE ERROR phrases' and 'Valid and invalid elementary moves')",
1445 basis: Basis::Documented,
1446 oracle: Oracle::EnterpriseCobol,
1447 },
1448 Assumption {
1449 id: CORRESPONDING_CHOICES,
1450 claim: "Where the Language Reference leaves CORRESPONDING open: pairs are processed in the order of the sending group's entries; the items of a FILLER group are not considered; an item is alphabetic when its PICTURE holds only A, as ironwork has no alphabetic category of its own; a numeric-edited item is not numeric for ADD and SUBTRACT, following the rule over the manual's example, which adds two; ADD and SUBTRACT evaluate every sending item before storing any receiver (C97), which differs from pair-by-pair only when a receiving item overlaps a later sending one; and no message is given when no items correspond",
1451 basis: Basis::Chosen,
1452 oracle: Oracle::EnterpriseCobol,
1453 },
1454 Assumption {
1455 id: STOP_LITERAL,
1456 claim: "STOP literal communicates the literal to the operator and suspends the program until the operator intervenes, then continues with the next statement (Language Reference for Enterprise COBOL 6.4, 'STOP statement'). ironwork has no operator to wait for: it writes the literal as DISPLAY does and continues at once",
1457 basis: Basis::Chosen,
1458 oracle: Oracle::EnterpriseCobol,
1459 },
1460 Assumption {
1461 id: NUMPROC_MIG_WARNS,
1462 claim: "NUMPROC(MIG) on a CBL or PROCESS card is a warning (W, return code 4), and the compile takes the default NUMPROC, NOPFD, as ironwork has no installation defaults, whatever NUMPROC an earlier option set (Migration Guide GC27-8715-03, Table 23, p. 112, and Table 32, p. 167). The guide gives neither the message's number nor its text: the message is ironwork's own",
1463 basis: Basis::Documented,
1464 oracle: Oracle::EnterpriseCobol,
1465 },
1466 Assumption {
1467 id: INVALID_OPTION_DISCARDED,
1468 claim: "A suboption that an option ironwork reads does not have, such as TRUNC(FAST), ARITH(X), NOCOMPILE(U) or a CODEPAGE that is not a number, is an error (E, return code 8) and the option is discarded, the setting before it staying in force, as the Migration Guide records for removed TEST suboptions: 'Error (Invalid option diagnostic, option discarded)' (GC27-8715-03, Table 34, p. 168); the Programming Guide shows the compiler diagnosing a CBL statement's options and carrying on (SC27-8714-03, pp. 279-280). The message's number and text are not in the manuals ironwork has: the text is ironwork's, unchanged from when the option stopped the compile. A CODEPAGE that is a number but no EBCDIC page ironwork carries, single-byte or one of the mixed pages of the Programming Guide's Table 47, still stops the compile (S), since IBM would compile the program in that page and ironwork cannot read it so; an option name that is in no table of IBM's still passes without a message",
1469 basis: Basis::Documented,
1470 oracle: Oracle::EnterpriseCobol,
1471 },
1472 Assumption {
1473 id: OPTIONS_WITHOUT_EFFECT,
1474 claim: "Options Enterprise COBOL 6.4 no longer has are accepted without effect. LIB, which the compiler now always behaves as having, and SIZE (Migration Guide GC27-8715-03, Table 32, p. 167) are informational (return code 0), as Enterprise COBOL 6.3 gives them for invocation parameters LIB and SIZE(2097152) in job output in the corpus: IGYOS4090-I 'The \"LIB\" option specification is no longer required. COBOL library processing is always in effect.' and IGYOS4013-I 'The \"SIZE\" option is no longer supported.' (SamMoussa961_COBOL, CLHELLO JOB03701); ironwork gives them for a CBL or PROCESS card too, and takes SZ as SIZE's abbreviation. FLAGSAA and NOFDUMP are warnings (W, return code 4), as the guide says IBM warns for each (Table 23, p. 112), with ironwork's text since the guide gives none. FDUMP, which IBM maps to TEST, and NOLIB pass without a message, as TEST does here; no source ironwork has shows IBM's message for NOLIB. The messages are ironwork's own words",
1475 basis: Basis::Observed,
1476 oracle: Oracle::EnterpriseCobol,
1477 },
1478 Assumption {
1479 id: NON_COBOL_CHARACTERS,
1480 claim: "A single-byte character outside the basic COBOL character set (Language Reference SC27-8713-03, Table 1, pp. 3-6), outside a literal, comment or PICTURE string, is accepted with IGYLI0163-E, 'Non-COBOL character \"%\" was found in column 8. The character was accepted.' (Migration Guide GC27-8715-03, p. 127), one message for each such character: an error (E, return code 8), so the program still runs under NOCOMPILE(S). Accepted means it is read as a character of the word it is in, or as a word of its own, which the parse then takes or refuses as it would any word. The guide calls non-COBOL the EBCDIC characters outside the set (p. 125), so control characters count; a character beyond U+00FF, which z/OS would hold in DBCS, is still refused, as are $ and &, whose own messages ironwork keeps. COPY REPLACING works on the text before it is read, so a character that REPLACING removes, which IBM diagnoses in the member, gives no message here",
1481 basis: Basis::Documented,
1482 oracle: Oracle::EnterpriseCobol,
1483 },
1484 Assumption {
1485 id: NO_PROGRAM_END,
1486 claim: "A program with no STOP RUN, GOBACK or EXIT PROGRAM statement anywhere in its PROCEDURE DIVISION, DECLARATIVES included, gets IGYPS2091-W, 'No \"STOP RUN\", \"GOBACK\" or \"EXIT PROGRAM\" was found in the program. Check program logic to verify that the program will exit.' (Migration Guide GC27-8715-03, p. 131): a warning (W, return code 4) with no line, in ironwork's words. A class definition and its methods, which end with EXIT METHOD, get no such warning. One that has an EXEC CICS RETURN or EXEC CICS XCTL is exempt by default, by the operator's choice pending an Enterprise COBOL listing of such a program, and --cics-return-warning says what it gets: once, the default, an informational note (return code 0) in place of the warning; always, the warning; never, nothing. The evidence points the other way: the CICS translator turns EXEC CICS RETURN into Call 'DFHEI1' using by content x'0e0800000600001000' end-call, with no GOBACK after it (CICS TS Application Programming Guide SC34-6433-06, pp. 87-88; IBM's CICS TS COBOL translation-output page shows the same), and the guide names only the three statements, so Enterprise COBOL itself would likely warn. The note is once per ironwork invocation because check, run and cics print the messages of one program, the first in the source",
1487 basis: Basis::Chosen,
1488 oracle: Oracle::EnterpriseCobol,
1489 },
1490 Assumption {
1491 id: USE_WITHOUT_PARAGRAPH,
1492 claim: "A DECLARATIVES section whose USE statement is followed at once by the next section or END DECLARATIVES gets IGYPS2036-I, 'A paragraph-name was missing after the \"USE\" statement.', informational (return code 0), on the line that follows, as Enterprise COBOL 6.3 gives it for CCVS85 IC401M, DB301M, DB302M and DB305M in the compile listings of eclipse-che4z's COBOL language server tests. No manual ironwork has lists the message. A USE statement followed by statements with no paragraph-name, and a debugging section read as a comment without WITH DEBUGGING MODE, get none, as no listing shows what IBM gives for them",
1493 basis: Basis::Observed,
1494 oracle: Oracle::EnterpriseCobol,
1495 },
1496 Assumption {
1497 id: PICTURE_ENDS_AT_ITS_SEPARATOR,
1498 claim: "A PICTURE character-string is delimited only by a separator space, comma, semicolon or period (Language Reference SC27-8713-03, p. 48), and a separator comma, semicolon or period is that character followed by a space (p. 50), so only the last such character before the space is a separator and any before it belong to the string: PIC 9,9,9,. is 9,9,9, with an insertion comma at its end, and PIC 999.. is 999. with its decimal point at its end, as CCVS85 NC125A writes them. As an IBM extension the string so ended may also be followed by a separator comma or semicolon and further clauses, where the 85 standard requires the separator period (p. 744), so PIC 999., VALUE ZERO is 999. too. DECIMAL-POINT IS COMMA exchanges the period's and comma's functions only within PICTURE strings and numeric literals (pp. 131, 208), not as separators, so the same holds under it",
1499 basis: Basis::Documented,
1500 oracle: Oracle::EnterpriseCobol,
1501 },
1502 Assumption {
1503 id: ZONED_COMPARED_AS_BYTES,
1504 claim: "A zoned integer compared with a nonnumeric operand, an alphanumeric, alphanumeric-edited, numeric-edited or group item, an alphanumeric or hexadecimal literal, or a figurative constant other than ZERO, is compared as the bytes it holds, without being read as a number: a numeric integer in such a comparison is treated as moved to an alphanumeric item of its size (Language Reference, comparison of numeric and alphanumeric operands), which for zoned data copies the digits. Under ZWB, IBM's default, a sign it overpunches is removed first (its zone made F); under NOZWB it is kept (Programming Guide SC27-8714-03, p. 431: 'Use NOZWB if you want to test input numeric fields for SPACES'); a separate sign is left out either way. So an unsigned item holding spaces equals SPACES, and a signed one does under NOZWB, where reading it as a number would end in a data exception. A scaled item is compared as its digits, as before",
1505 basis: Basis::Chosen,
1506 oracle: Oracle::EnterpriseCobol,
1507 },
1508 Assumption {
1509 id: APOST_EVERYWHERE,
1510 claim: "Under APOST the figurative constant QUOTE is the apostrophe wherever the program uses it: X'7D' in an alphanumeric item, X'0027' in a national one, and as an entry of an ALPHABET clause. The Programming Guide says only that [ALL] QUOTE and QUOTES represent apostrophes under APOST and quotation marks under QUOTE, and that either may delimit a literal whichever is in effect (SC27-8714-03, p. 347)",
1511 basis: Basis::Chosen,
1512 oracle: Oracle::EnterpriseCobol,
1513 },
1514 Assumption {
1515 id: CURRENCY_OPTION,
1516 claim: "CURRENCY(literal), in a program with no CURRENCY SIGN clause of its own or of a containing program, acts as a CURRENCY SIGN clause whose value is the literal's one character, standing for itself: that character is the PICTURE currency symbol in place of $, and an edited item shows it. The Language Reference says the currency symbol is $ or the one character the CURRENCY option or a CURRENCY SIGN clause gives, and that a CURRENCY SIGN clause makes the option ignored (SC27-8713-03, pp. 130, 211); the Programming Guide lists the characters the literal may not be (SC27-8714-03, p. 358). That the currency sign value is the character itself, and that nothing is said when the option is ignored, are chosen. A hexadecimal literal is read in the program's code page once every card is applied, so CODEPAGE may follow it; one whose character the option may not name is an error and the option is discarded, as an invalid suboption is (C121). NUMVAL-C's default currency string is not changed here",
1517 basis: Basis::Chosen,
1518 oracle: Oracle::EnterpriseCobol,
1519 },
1520 Assumption {
1521 id: NSYMBOL_DBCS,
1522 claim: "Under NSYMBOL(DBCS) an N literal is a DBCS literal and a PICTURE of N, alone or with B, and no USAGE, its own or a group's, is USAGE DISPLAY-1 (Programming Guide SC27-8714-03, pp. 387-388); the option is read from the CBL and PROCESS cards before the source is read, as a card is the only place a program can set it. NX literals are not N literals. NSYMBOL(NATIONAL) with NODBCS on the cards, which are one level of precedence, leaves DBCS in effect, as Table 46 forces it (p. 344), with a warning (W, return code 4) in ironwork's words, since the message Enterprise COBOL gives for an option dropped in conflict resolution is IGYOS4020-W (J19), as for INITIAL with THREAD (C217); NODBCS alone, with NSYMBOL(NATIONAL) only as the default, is taken as written",
1523 basis: Basis::Chosen,
1524 oracle: Oracle::EnterpriseCobol,
1525 },
1526 Assumption {
1527 id: INITIAL_UNDER_THREAD,
1528 claim: "The INITIAL option with THREAD, on the same or different CBL and PROCESS cards and in either order, is ignored and NOINITIAL forced, with an error message (Programming Guide SC27-8714-03, p. 344, Table 46); under THREAD IBM diagnoses the INITIAL option as an error (p. 418). ironwork gives a warning (W, return code 4) in its own words, as for NORENT with THREAD, since the message Enterprise COBOL gives for an option dropped in conflict resolution is IGYOS4020-W (J19), and the program, its nested programs and the options in its load module are NOINITIAL. An IS INITIAL clause under THREAD keeps J13's error. A class definition gets the warning once and its methods none; the option makes no method initial, as INITIAL is an attribute of a program and its nested programs (p. 374; Language Reference SC27-8713-03, p. 103)",
1529 basis: Basis::Chosen,
1530 oracle: Oracle::EnterpriseCobol,
1531 },
1532 Assumption {
1533 id: VLR_WITHOUT_VARYING,
1534 claim: "VLR(STANDARD) checks the length of a variable-length record a READ returns against the least and greatest of the file's level-01 records, and VLR(COMPAT) against its RECORD IS VARYING IN SIZE FROM min TO max (Programming Guide SC27-8714-03, pp. 422-424, Table 52); a FROM or TO the clause leaves out is the least or greatest level-01 record (Language Reference SC27-8713-03, p. 187), and a level-01 record's least length counts an OCCURS DEPENDING ON table at its fewest occurrences, 1 when the entry has no integer-1 TO (pp. 188, 204). A file without RECORD IS VARYING is checked against its level-01 records under COMPAT as under STANDARD, whether it has RECORD CONTAINS integer-4 TO integer-5, whose integers the record descriptions decide and must match (pp. 187, 191), RECORD CONTAINS integer-3, or no RECORD clause: the guide names only the VARYING declaration as what COMPAT checks, and gives no case without one. A file with no level-01 record is checked against its record area alone",
1535 basis: Basis::Chosen,
1536 oracle: Oracle::EnterpriseCobol,
1537 },
1538 Assumption {
1539 id: VLR_RECORDS_CHECKED,
1540 claim: "VLR changes the file status alone: under either setting the READ succeeds and delivers the same bytes (Programming Guide SC27-8714-03, p. 422). A record shorter than the check's minimum fills the record area only as far as its length, leaving the rest as it was, which IBM calls undefined, and one longer than the record area is truncated to it (Language Reference SC27-8713-03, pp. 431, 434); the record area, the larger of the RECORD clause's maximum and the longest level-01 record, is taken as the 'maximum record definition size' (p. 431), so a 70-byte record of Table 52's file is delivered whole and READ INTO moves its 70 bytes (p. 188). The check covers records of variable length: a sequential file whose DD or FD is variable, and an indexed, relative or I-O sequential file held in that format. A fixed-length record keeps its status 04 for a data set that ends in a short record or a record longer than the area, whatever VLR says, and a line-sequential file or text DD, whose short lines IBM fills with spaces (Programming Guide p. 218), is not checked",
1541 basis: Basis::Chosen,
1542 oracle: Oracle::EnterpriseCobol,
1543 },
1544 Assumption {
1545 id: VSAM_DATA_SET_LEFT_OPEN,
1546 claim: "A VSAM data set, which an indexed or relative file's DD names, keeps the open-for-output indicator of its catalog entry as a file beside it, its path with .open-for-output added: OPEN OUTPUT, I-O or EXTEND sets it, and only the successful CLOSE after an OPEN for output takes it away, so an OPEN INPUT leaves it as it was (z/OS 3.1 DFSMS Using Data Sets, idad400/d4011). At OPEN, VSAM implicitly issues a VERIFY when it finds the indicator on, and the OPEN goes on (d4011, idad400/verif; DFSMS Macro Instructions for Data Sets, idad500/x1cb, reason code 118, X'76', the high-used RBA verified). ironwork's verify always succeeds, and the OPEN's status is 97 under VSAMOPENFS(COMPAT), the default, or 00 under VSAMOPENFS(SUCC) (Programming Guide SC27-8714-03, pp. 199, 424; Language Reference SC27-8713-03, p. 303, Table 34), for an OPTIONAL file too. Status 97 sets file status key 1 to 9, so it runs the file's EXCEPTION/ERROR procedure (LR p. 706), and it reports an OPEN that succeeded, so with no FILE STATUS and no procedure the run goes on (LR p. 411; PG p. 204). Files a run unit leaves open are closed when it ends, and an abend leaves the indicator on only where Language Environment closes no file (C152). Sequential files, which ironwork does not hold as VSAM data sets, are never marked; nor are the files of CICS file control, which opens and closes them with each task. ironwork job deletes the mark with its data set, DEFINE CLUSTER starts a data set without one, and REPRO into a data set takes it away, as the CLOSE after REPRO's OPEN for output does",
1547 basis: Basis::Chosen,
1548 oracle: Oracle::EnterpriseCobol,
1549 },
1550 Assumption {
1551 id: DISPSIGN_SEPARATE,
1552 claim: "Under DISPSIGN(SEP) DISPLAY shows a signed binary, packed or overpunched zoned item as a leading + or - followed by the digits DISPSIGN(COMPAT) shows, unsigned (Programming Guide SC27-8714-03, pp. 362-363, Table 48), so a COMP-5 or TRUNC(BIN) binary item keeps its 5, 10, 19 or 20 digits (C14). The Language Reference shows the sign 'as if SIGN IS SEPARATE was specified' (SC27-8713-03, p. 334), which without LEADING would put it last; the guide's leading sign and its table are followed. IBM's table lists no other kind, and the rest is chosen: a zoned item with SIGN SEPARATE, whose sign is separate already, shows as stored, a trailing sign staying last; an unsigned item, and a value that is not an item (a literal, a function's result), is unchanged; a national decimal item, which ironwork does not lay out yet, would be unchanged too, IBM naming only binary, packed and zoned. An overpunched zoned item is read as its bytes, as COMPAT shows them, not as arithmetic reads it: the sign is - when the sign position's zone is X'B' or X'D' and + for any other zone, one that is no sign included, and the digits are the bytes with that zone made X'F', so invalid data is shown rather than ending the run",
1553 basis: Basis::Chosen,
1554 oracle: Oracle::EnterpriseCobol,
1555 },
1556 Assumption {
1557 id: LILIAN_INTEGER_DATES,
1558 claim: "Under INTDATE(LILIAN) every integer date a date function takes or returns is a Lilian day, 15 October 1582 being day 1 (Programming Guide SC27-8714-03, pp. 59, 375): INTEGER-OF-DATE, DATE-OF-INTEGER, DAY-OF-INTEGER, INTEGER-OF-DAY, FORMATTED-DATE, FORMATTED-DATETIME and INTEGER-OF-FORMATTED-DATE, which the Language Reference marks as INTDATE's (SC27-8713-03, pp. 551, 555, 571, 573, 583, 585, 587). The Language Reference gives the ranges under ANSI only: integer dates 1 to 3,067,671 and years 1601 to 9999 (p. 509). Under LILIAN the dates run, as Language Environment's date services take them, from 15 October 1582 to 31 December 9999, so integer dates from 1 to 3,074,324, standard dates from 15821015, Julian dates from 1582288, and a formatted date's year from 1582. TEST-DATE-YYYYMMDD, TEST-DAY-YYYYDDD, TEST-FORMATTED-DATETIME and SECONDS-FROM-FORMATTED-TIME carry no INTDATE note and keep years from 1601 (pp. 637, 653, 655, 657). FORMATTED-CURRENT-DATE takes and gives no integer date, so INTDATE does not touch it. COMBINED-DATETIME, the other function the note is on (p. 541), takes an integer date in the same range",
1559 basis: Basis::Chosen,
1560 oracle: Oracle::EnterpriseCobol,
1561 },
1562 Assumption {
1563 id: CEECBLDY_UNDER_LILIAN,
1564 claim: "Under INTDATE(LILIAN) a CALL whose target is the literal 'CEECBLDY' is diagnosed and calls CEEDAYS, which takes the same arguments and gives a Lilian day (Programming Guide SC27-8714-03, p. 375). The guide gives neither the message nor its severity: it is a warning (W, return code 4) in ironwork's words, since the program no longer calls what it names. A CALL of an identifier that holds 'CEECBLDY' is not converted, as the guide names the literal only, and ends the run as ironwork does not provide CEECBLDY; nor is a literal under INTDATE(ANSI)",
1565 basis: Basis::Chosen,
1566 oracle: Oracle::EnterpriseCobol,
1567 },
1568 Assumption {
1569 id: COMPLETE_SET_OF_QUALIFIERS,
1570 claim: "Under QUALIFY(EXTEND) a reference the standard's rules find ambiguous names the one candidate it gives a complete set of qualifiers, 'every level in the containing hierarchy of names' (Programming Guide SC27-8714-03, p. 400; 'every qualifier is specified', Language Reference SC27-8713-03, p. 68); with no such candidate, or two, it stays ambiguous. A data item's complete set is the name of every group that holds it, nearest first, up to its level-01 entry: a FILLER or unnamed group has no name to give and is no level of it, and a level-66 item's hierarchy is its record. A record's file-name may follow but is not needed, since the Language Reference lets a level-01 name that is the only one of its level be referenced under EXTEND (p. 67), which a record of an FD or SD could not be if its file-name were needed (p. 69); a LINAGE-COUNTER's complete set is its file-name or nothing. A condition-name's hierarchy starts with its conditional variable, which qualifies it (p. 70). The rule applies to RENAMES operands too, among the items of the record. A SUM operand naming a REPORT SECTION entry is found by its report-name and does not follow it",
1571 basis: Basis::Chosen,
1572 oracle: Oracle::EnterpriseCobol,
1573 },
1574 Assumption {
1575 id: INSPECT_FUNCTION_SUBJECT,
1576 claim: "INSPECT's identifier-1 can be an alphanumeric or national function-identifier in a TALLYING-only INSPECT (format 1): a function-identifier can be used wherever a sending data item of its category can, except as a receiving operand (Language Reference SC27-8713-03, p. 77), the INSPECT data flow evaluates a function-identifier once, as the first operation (p. 360), and TALLYING leaves identifier-1 unchanged. REPLACING (formats 2 and 3) and CONVERTING (format 4) copy their result back to identifier-1 (Table 40, p. 359), so a function there is a receiving operand and is refused when compiled (S); so is an integer or numeric function as identifier-1, since one can be used only where an arithmetic expression can (pp. 77, 505) and identifier-1 must be a DISPLAY, DISPLAY-1 or NATIONAL item or group (p. 355). MIN, MAX and CONTENT-OF, whose type follows their arguments, are not refused. The manuals give neither the message numbers nor their text: the messages are ironwork's. A function-identifier as identifier-3 to identifier-7 is read as any sending operand is",
1577 basis: Basis::Documented,
1578 oracle: Oracle::EnterpriseCobol,
1579 },
1580 Assumption {
1581 id: INSPECT_NATIONAL_FUNCTION_RESULT,
1582 claim: "INSPECT TALLYING of a national function result counts national characters (two-byte encoding units) and matches comparands and BEFORE or AFTER INITIAL delimiters only at character boundaries, and a figurative constant there is one national character (Language Reference SC27-8713-03, p. 355). An alphanumeric literal or value there is made national as C232 says; an inspected data item of usage NATIONAL is C230. No intrinsic function returns DBCS, so a DBCS function result does not arise",
1583 basis: Basis::Documented,
1584 oracle: Oracle::EnterpriseCobol,
1585 },
1586 Assumption {
1587 id: NATIONAL_CASE_AND_REVERSE,
1588 claim: "UPPER-CASE, LOWER-CASE and REVERSE of a national argument are national (Language Reference SC27-8713-03, pp. 597, 635, 671), as TRIM's already was (p. 665), and lowering types all four so. REVERSE keeps a surrogate pair as one character (p. 635). The case functions map each character by Unicode's case mapping where that gives one character and leave it alone otherwise, so the result keeps the argument's length as the manual requires; IBM names UnicodeData.txt, whose simple mappings differ from this in a few characters, such as U+0130, which is left as it is here",
1589 basis: Basis::Chosen,
1590 oracle: Oracle::EnterpriseCobol,
1591 },
1592 Assumption {
1593 id: INVDATA_CLEANSIGN,
1594 claim: "Under INVDATA(CLEANSIGN), the default once INVDATA is given and part of ZONEDATA(MIG) and ZONEDATA(NOPFD), a packed-decimal or zoned item whose sign half-byte is a digit, 0 to 9, is read with that half-byte made F, positive, rather than ending in a data exception: the compiler 'generates code to clean the sign nibble of USAGE DISPLAY and USAGE PACKED-DECIMAL data items on input to compare, add, subtract, multiply, and divide operations', and not for SIGN IS SEPARATE (Programming Guide SC27-8714-03, p. 378). Which valid sign the cleaning produces is not stated; F is chosen. It applies wherever the item is read as a number, a MOVE's sending item included, where IBM names only comparisons and arithmetic. Under NOCLEANSIGN and NOINVDATA such a sign is a data exception, as before",
1595 basis: Basis::Chosen,
1596 oracle: Oracle::EnterpriseCobol,
1597 },
1598 Assumption {
1599 id: INVDATA_ZONES_COMPARED,
1600 claim: "Under INVDATA(NOFORCENUMCMP), the default once INVDATA is given and part of ZONEDATA(NOPFD), an unsigned zoned integer compared with zero (ZERO, or a numeric literal of value zero), or with an unsigned zoned integer of its own length, is compared as the bytes it holds, zones included, so an item holding X'F0F040F0' is not equal to ZERO: the compiler compares zoned data 'in the same manner as COBOL 4 or earlier versions', by an alphanumeric comparison where those considered the zone bits, and IBM's VALUE1 example gives false under INVDATA(NOFORCENUMCMP) at any OPT setting (Programming Guide SC27-8714-03, pp. 377-378). Which comparisons COBOL 4 made by their bytes is not listed; these two, where the bytes of equal values are always equal, are chosen, and a condition-name's value is compared as the relation of the two would be. Other comparisons, and every comparison under FORCENUMCMP and ZONEDATA(MIG), read the digits and ignore the zones; NOINVDATA is C262",
1601 basis: Basis::Chosen,
1602 oracle: Oracle::EnterpriseCobol,
1603 },
1604 Assumption {
1605 id: OPTIMIZED_ZONES_COMPARED,
1606 claim: "Under NOINVDATA, IBM's default, the comparisons of C223 (an unsigned zoned integer against zero or against an unsigned zoned integer of its own length, a condition-name's values included) are made by their bytes, zones included, at OPTIMIZE(1) and OPTIMIZE(2), and as numbers at OPTIMIZE(0), IBM's default: NOINVDATA lets 'the compiler ... generate a string comparison to avoid numeric conversion', and IBM's VALUE1 example, X'F0F040F0' compared with ZERO, is true at OPT(0) and false at OPT(1) and OPT(2) (Programming Guide SC27-8714-03, pp. 377-378). Every other comparison reads its operands' digits at any level, so a digit half above 9, or a sign the NUMPROC setting does not accept, is a data exception at the comparison, as PACK and CP give it; that this is IBM's code for them at OPT(1) and OPT(2) is recalled, not documented. IBM states only that invalid data makes a reference 'undefined' and its results 'unpredictable' (p. 53), that 'digits and sign codes must be valid no matter what options are used' and data that is not may 'behave differently at different levels of optimization' (p. 395), and that NOCLEANSIGN increases 'the probability of a S0C7 abend' when an operand of a comparison has an invalid sign (p. 378). A numeric literal of value zero is compared as ZERO is, which the optimizer cannot tell apart; that too is chosen. Under NOINVDATA and --dialect gnucobol the comparison with zero is made as numbers at every OPTIMIZE level, and the comparison of two unsigned zoned integers of one length by their bytes at every level, as cobc compares them",
1607 basis: Basis::Chosen,
1608 oracle: Oracle::EnterpriseCobol,
1609 },
1610 Assumption {
1611 id: ALPHANUMERIC_MOVED_UNCHECKED,
1612 claim: "An alphanumeric sender MOVEd to a zoned or packed integer item without P scaling is not checked for digits at the MOVE: the receiver gets the low half of each of the sender's last bytes as its digits, zeros to the left, stored as a positive value, and a byte whose low half is not a digit (an asterisk, X'5C') leaves that half in the receiver, so the data exception comes where the item is next read as a number. The Language Reference treats such a sender as an unsigned numeric integer (MOVE statement, elementary moves); the instructions a MOVE compiles to, a byte copy, PACK and UNPK, raise no data exception (z/Architecture Principles of Operation, PACK and UNPACK), and the decimal instructions of arithmetic do. A receiver with decimal places or P scaling, a binary receiver (CVB checks digits) and a numeric-edited one still read the sender as a number at the MOVE",
1613 basis: Basis::Recalled,
1614 oracle: Oracle::EnterpriseCobol,
1615 },
1616 Assumption {
1617 id: NUMERIC_MOVED_UNCHECKED,
1618 claim: "A zoned or packed sender is not checked at a MOVE, or at the MOVE of a WRITE, REWRITE or RELEASE FROM phrase, whose generated code checks nothing: to a zoned item, to a packed item from a zoned one, to a packed item of its own kind and scaling under NUMPROC(PFD), and to an alphanumeric, alphanumeric-edited or group item. Those compile to a byte copy, PACK, UNPK and the OI that makes a sign F, which raise no data exception (z/Architecture Principles of Operation, PACK and UNPACK); NUMCHECK(ZON(LAX)) likewise leaves the sender of a zoned-to-zoned or zoned-to-alphanumeric MOVE unchecked and checks it 'if the sender is subsequently used in a numeric context' (Programming Guide SC27-8714-03, p. 391), and IBM leaves what invalid data gives to the generated code, which differs with OPT and ARCH (Migration Guide GC27-8715-03, pp. 201, 205-206). Where the sender's digits or sign are not decimal, each receiver digit takes the low half of the sender's digit of the same power of ten, zero where it has none, so a non-digit stays in the receiver and the data exception comes where the item is next read as a number; a sign half that is a digit stays in a signed receiver's sign place, an unsigned receiver's sign is F, and SIGN SEPARATE reads any character but '-' as positive. An alphanumeric receiver gets a zoned sender's bytes, an overpunched sign's zone made F, or a packed sender's digits unpacked with F zones. A packed sender to another packed shape (ZAP or SRP), and any zoned or packed sender to a binary (CVB), numeric-edited (ED) or floating-point receiver, is still read as a number at the MOVE and ends in S0C7 there. A sender whose digits and sign are decimal moves as before. Which instructions IBM generates for each pair is recalled, not documented",
1619 basis: Basis::Recalled,
1620 oracle: Oracle::EnterpriseCobol,
1621 },
1622 Assumption {
1623 id: STATEMENT_LIMIT_IS_TIME,
1624 claim: "A run given a statement limit (run or job --statement-limit) ends with S322 once that many statements have started, at the next start of the first statement of the loop it is in (the lowest-placed of the statements that recur among the last 4,096 starts, in their outermost frame and the program's own source), or where the count ran out when no statement recurs or that one does not start again within 4,096 more; S322 stands for the system completion code z/OS gives a job step that runs past the CPU time its JOB or EXEC TIME= parameter allows (MVS System Codes, 322). A count of statement starts stands in for CPU time so the end falls at the same statement on every run and on both executors, and placing it at the loop rather than where the count ran out keeps it there whatever ran before the loop; it says nothing of how long the program would run on z/OS, and a loop whose iterations start no statement is not stopped by it",
1625 basis: Basis::Chosen,
1626 oracle: Oracle::EnterpriseCobol,
1627 },
1628 Assumption {
1629 id: INITCHECK_ANALYSIS,
1630 claim: "INITCHECK follows the procedure's control flow to a fixed point: IF, EVALUATE, SEARCH, the ON and NOT ON phrases, inline and out-of-line PERFORM, GO TO, GO TO DEPENDING ON, ALTER, NEXT SENTENCE, EXIT PARAGRAPH, SECTION, PERFORM and PERFORM CYCLE, STOP RUN and GOBACK, and fall-through between paragraphs; under LAX an item counts as set where some path to the statement sets it, under STRICT where every path does (Programming Guide SC27-8714-03, pp. 373-374). As IBM states: LINKAGE, FILE SECTION, EXTERNAL and GLOBAL items are not analysed; a table is one item, set when any element is; an item referenced anywhere with reference modification counts as set; a BY REFERENCE argument is not a use and a BY CONTENT or BY VALUE one is; an item whose address is taken (ADDRESS OF, a BY REFERENCE argument, an EXEC SQL host variable, an EXEC CICS argument) makes its whole level-01 record, with any record that REDEFINES it, address-taken, and every CALL, INVOKE and EXEC statement sets all address-taken records; setting through a pointer is not tracked. What else sets an item is chosen: a VALUE clause; a receiver of MOVE, COMPUTE, ADD, SUBTRACT, MULTIPLY, DIVIDE (and REMAINDER), INITIALIZE, ACCEPT, SET (a condition-name sets its conditional variable), STRING, UNSTRING (with DELIMITER IN, COUNT IN, POINTER and TALLYING), INSPECT REPLACING or CONVERTING and an INSPECT TALLYING counter, SEARCH VARYING, PERFORM VARYING, READ INTO, RETURN INTO, JSON PARSE INTO, JSON and XML GENERATE's receiver and COUNT IN, and a CALL's or INVOKE's RETURNING item; a file's FILE STATUS data-names and RELATIVE KEY by any statement on the file; SQLCA, SQLDA and the DL/I interface block, which the coprocessor and translators pass on every EXEC statement, and a symbolic map RECEIVE MAP or SEND MAP names, are address-taken. The records of the CICS, SQL and DL/I members ironwork supplies when no library holds them (DFHAID, DFHBMSCA, SQLCA and the like) are not analysed: IBM's copies give constants VALUE clauses that ironwork's leave out, and the translators set the rest. Setting a group sets every elementary item in it and a group is set when they all are; setting an item sets every item whose storage it shares through REDEFINES or RENAMES. A use reads each elementary item of a group, and a group is reported by its first that is not set. FILLER items are not analysed. An out-of-line PERFORM, a SORT's input or output procedure and XML PARSE's processing procedure are calls of their range, which returns at the end of its last paragraph: a range's entry state joins those of every statement that performs it, and each continues with its own state plus what every path through the range sets. PERFORM ... UNTIL or VARYING with TEST BEFORE, PERFORM ... TIMES with other than a literal above zero, and XML PARSE's procedure may run no times; WHEN OTHER aside, EVALUATE may select no WHEN. AT END, INVALID KEY and ON EXCEPTION paths leave unset what the statement would otherwise set, and so does ON SIZE ERROR; without such a phrase only the successful path continues. Every operand of a condition is a use, short-circuiting aside, and FUNCTION LENGTH's argument is not. A USE FOR DEBUGGING section may run as control enters each procedure it names, or any procedure for ALL PROCEDURES, and a USE AFTER EXCEPTION/ERROR section after any statement on a file it serves, or on any file for an open mode; procedures only EXEC CICS HANDLE, EXEC SQL WHENEVER or USE BEFORE REPORTING would reach, and code no path reaches, are not analysed. An ENTRY point starts from the VALUE clauses. IBM's analysis is more accurate under OPT(1) and OPT(2) (p. 374), where its optimizer may prove paths impossible that this one does not",
1631 basis: Basis::Chosen,
1632 oracle: Oracle::EnterpriseCobol,
1633 },
1634 Assumption {
1635 id: INITCHECK_MESSAGE,
1636 claim: "INITCHECK's message is IBM's 7311, severity W, return code 4: the Programming Guide's sample MSGEXIT names message 7311(W) as 'the case of INITCHECK messages about uninitialized data items' (SC27-8714-03, Appendix E, p. 841). Its prefix and text are not in the manuals ironwork holds, so ironwork gives the warning in its own words: once per statement and item used, at the statement's position, naming the item and, for a group, the first of its elementary items not set, as IBM reports only the first uninitialized item in a group (p. 374). Under LAX it says no path to the statement sets the item; under STRICT that a path does not",
1637 basis: Basis::Chosen,
1638 oracle: Oracle::EnterpriseCobol,
1639 },
1640 Assumption {
1641 id: PARMCHECK_BUFFER,
1642 claim: "Under PARMCHECK the buffer of n bytes, 100 by default, starts at the byte after the end of the WORKING-STORAGE the program declares, the furthest any of its 01 and 77 items reaches, with no slack bytes between, so a CALL that writes one byte past the last item changes it: IBM puts it 'following the last data item in the WORKING-STORAGE section' (Programming Guide SC27-8714-03, p. 397) and does not say whether it is aligned; were it on a doubleword, a write into the slack bytes before it would go unseen on z/OS. The special registers and Report Writer data ironwork adds to WORKING-STORAGE after the program's own items (the SORT, XML and JSON registers, DEBUG-ITEM, LINAGE-COUNTER, a report's counters), which are no items the program declares, follow the buffer, as do the file record areas and index-names; every offset after the buffer moves by n, and a program compiled without PARMCHECK keeps its layout byte for byte. The buffer is no data item and nothing names it. It holds zeros until a CALL sets it to X'AA'; each CALL of a program, of a Language Environment callable service and through a function-pointer sets it after the arguments are evaluated and checks it once the called program returns, before RETURNING is stored and before NOT ON EXCEPTION runs. A CALL after which the run ends, by STOP RUN or an abend in the called program, is not checked",
1643 basis: Basis::Chosen,
1644 oracle: Oracle::EnterpriseCobol,
1645 },
1646 Assumption {
1647 id: PARMCHECK_MESSAGE,
1648 claim: "IBM documents what PARMCHECK's message holds but not its number or text: under MSG a runtime warning with the name of the parameter, the line number of the CALL statement and the program name, issued after the CALL, and under ABD a similar message at a terminating level that causes an abend (Programming Guide SC27-8714-03, p. 397); the corpus of public job output searched has PARMCHECK only in option listings. The message is ironwork's, with no IGZ number: 'PARMCHECK: SUB, called at line 14 of program MAIN, wrote past the end of WORKING-STORAGE, beyond parameter LAST-ITEM', the program being the one with the CALL. Under MSG it goes to standard error after 'ironwork: line:column: ' and the run goes on; under ABD the run ends with U4038, as a Language Environment condition of severity 3 that nothing handles ends it under the default ABTERMENC(ABEND), as an SSRANGE failure does. The parameter named is chosen: of the CALL's arguments whose storage starts in the calling program's own WORKING-STORAGE, so BY REFERENCE ones (a BY CONTENT or BY VALUE argument is a copy elsewhere), a LINKAGE item counting when its address is there, the one starting nearest the buffer, since a called program that declares it longer reaches the buffer soonest; of two starting at the same byte, the later in the USING list. With no such argument the message names none. Which bytes changed is not reported",
1649 basis: Basis::Chosen,
1650 oracle: Oracle::EnterpriseCobol,
1651 },
1652 Assumption {
1653 id: NUMCHECK_SENDERS,
1654 claim: "NUMCHECK (and ZONECHECK, which is NUMCHECK(ZON)) tests an item where a statement reads it as a sender: an operand of an arithmetic expression, a comparison or a condition-name's variable, a MOVE's sender, a receiver that is also a sender (ADD A TO B), a subscript, and a BY CONTENT or BY VALUE argument, before the statement uses it; not a receiver alone, a BY REFERENCE argument, a class test, or DISPLAY, which shows the bytes and uses no value (Programming Guide SC27-8714-03, pp. 388-391, 427). A zoned item is tested as IF NUMERIC tests it, after INVDATA(CLEANSIGN) has cleaned its sign (C222); a packed one too, with the spare half-byte of an even digit count zero; a binary one for a value of more digits than its PICTURE, except COMP-5, and under TRUNC(BIN) only with BIN(TRUNCBIN). An alphanumeric item moved to a numeric receiver is tested as an unsigned integer. ZON(NOALPHNUM) leaves a zoned item untested in a comparison with an alphanumeric item, literal or figurative constant. ZON(LAX) leaves a zoned item moved to a zoned or alphanumeric receiver untested, and tolerates the two redefinitions of C280. A test the compiler finds always fails is reported when compiled and removed (C281). Each test is made each time the statement runs, where IBM removes redundant ones, so a loop may report more often than Enterprise COBOL does",
1655 basis: Basis::Chosen,
1656 oracle: Oracle::EnterpriseCobol,
1657 },
1658 Assumption {
1659 id: NUMCHECK_MESSAGE,
1660 claim: "NUMCHECK's run-time message names the item, its bytes in hexadecimal, the program and, by its position, the line, as IBM's does (Programming Guide SC27-8714-03, p. 391), in ironwork's words with no IGZ message number, since no Enterprise COBOL output ironwork has shows one. Under MSG it is written to the error stream and the statement runs; under ABD the run ends with U4038, the abend a Language Environment condition of severity 3 gives, as SSRANGE's does",
1661 basis: Basis::Chosen,
1662 oracle: Oracle::EnterpriseCobol,
1663 },
1664 Assumption {
1665 id: PARM_ARGUMENTS_BEFORE_LAST_SLASH,
1666 claim: "A COBOL main program that EXEC PGM=,PARM= starts receives as its program arguments what precedes the PARM's last slash, the rest being runtime options; when there are only invalid runtime options the whole string is the argument, so 11/16/1967 reaches the program whole. This is CBLOPTS(ON), the non-CICS default CBLOPTS=((ON),OVR) (Language Environment Programming Guide, COBOL compatibility considerations; Programming Reference, CBLOPTS)",
1667 basis: Basis::Documented,
1668 oracle: Oracle::EnterpriseCobol,
1669 },
1670 Assumption {
1671 id: PARM_AREA_PADDED,
1672 claim: "A runtime option after the last slash is known by its full name, or its NO form, from the CEEXOPT sample's list; an abbreviation counts as invalid, so a PARM whose only runtime options are abbreviated reaches the program whole. The first PROCEDURE DIVISION USING item addresses a halfword length and the arguments in the program's code page, followed by X'00' up to 100 bytes, JCL's longest PARM, so a program that reads its whole PIC X(100) parameter field reads zeros past the arguments rather than leaving storage. A step with no PARM passes a length of zero. Items after the first are not addressed",
1673 basis: Basis::Chosen,
1674 oracle: Oracle::EnterpriseCobol,
1675 },
1676 Assumption {
1677 id: ABBREVIATED_RELATIONS,
1678 claim: "An abbreviated combined relation condition (Language Reference SC27-8713-03, pp. 287-289) implies the last stated subject, or the last stated subject and relational operator, a NOT just before the operator being part of it, so A NOT > B OR C is (A NOT > B) OR (A NOT > C), and IS may come before an implied operator, as CCVS85 NC250A writes AND IS NOT LESS THAN. Parentheses after AND, OR or NOT take the implied subject and operator in. After the right parenthesis the subject and operator stated before the parentheses are current again, whatever was stated inside, so A = B AND (C OR < D) OR 2 ends with A = 2; the manual's rule 10 does not say which operator follows a right parenthesis. A class or sign condition leaves the implied subject and operator as they were. A bare name after AND or OR, with NOT or left parentheses between or not, is a condition-name when it names one and otherwise an object, decided when the name is resolved. A relational operator followed by parentheses that hold objects joined by AND, OR and NOT is distributed over them (rule 5), and NOT just after that parenthesis is refused. The manual's examples (Table 31, p. 289) agree with their unabbreviated forms in ironwork's tests; no Enterprise COBOL listing ironwork has shows what follows a right parenthesis or a class condition in an abbreviation",
1679 basis: Basis::Chosen,
1680 oracle: Oracle::EnterpriseCobol,
1681 },
1682 Assumption {
1683 id: PARAGRAPH_IN_OWN_SECTION,
1684 claim: "A paragraph-name need not be qualified when referred to within the section in which it appears (Language Reference SC27-8713-03, p. 68), so an unqualified name that more than one procedure has names the paragraph of that name in the section the reference is written in, when that section has one; from any other section it is refused as naming more than one paragraph, as before. The manual states the rule for references in general; ironwork applies it to each. Before any check, the compiler qualifies the names of GO TO, GO TO DEPENDING ON, PERFORM and its THRU, ALTER, SORT and MERGE procedures and XML PARSE's processing procedure with the section they are written in, so the compiler, the interpreter and the LIR lowering find the same paragraph. USE FOR DEBUGGING ON resolves from its declarative section, EXEC SQL WHENEVER GO TO from the paragraph of the SQL statement it follows, as the precompiler writes its GO TO there, and an EXEC CICS HANDLE label from the paragraph of the HANDLE command, with the same rule. CCVS85 NC208A's GO TO PAR-3C, unqualified in the section that has a PAR-3C, is the corpus case",
1685 basis: Basis::Chosen,
1686 oracle: Oracle::EnterpriseCobol,
1687 },
1688 Assumption {
1689 id: REPLACE_STATEMENT,
1690 claim: "REPLACE applies after COPY, to the whole expanded text, EXEC SQL and EXEC CICS statements included, as the Db2 coprocessor has it (Language Reference SC27-8713-03, pp. 708-712; Programming Guide SC27-8714-03, p. 513). A REPLACE is in effect from its period to the next REPLACE statement or the end of the source file: the manual says the end of the separately compiled program, so a batch of several programs in one file carries a REPLACE past the END PROGRAM of the one it is in, where IBM would stop it. The word REPLACE starts a statement only when pseudo-text, LEADING, TRAILING or OFF follows it, and wherever it is: the manual asks for a separator period before it, but its own example (pp. 709-710) has one after a DISPLAY with none, and gives that program's output. Matching is COPY REPLACING's: text words compared a word at a time, case aside outside literals, with a separator comma or semicolon never a word, so pseudo-text-1 that is only a comma or semicolon matches nothing. The replacing text takes the position of the first word it replaces, for messages and for Area A, where IBM puts each of its words in the area it is written in within pseudo-text-2; and REPLACE ALSO and REPLACE LAST OFF, which the 2014 standard has and the manual does not, are refused",
1691 basis: Basis::Chosen,
1692 oracle: Oracle::EnterpriseCobol,
1693 },
1694 Assumption {
1695 id: VARIABLY_LOCATED_ITEMS,
1696 claim: "An item after an OCCURS DEPENDING ON table in its record, and not within it, is variably located (Language Reference SC27-8713-03, pp. 205-206; Programming Guide SC27-8714-03, pp. 81-82): each reference places it back from where the table's maximum puts it by the bytes of the occurrences past the object's current value, and a group holding such tables is as long as their current counts make it; its VALUE is placed as if each table held its maximum (LR p. 246). A receiving group that holds the objects of its tables is at its maximum length only when nothing after it in its record moves with them, since the guide has the actual length used when a variably located item follows (PG pp. 78-79). The location is worked out from the objects' values at each reference, so a count changed between two references moves what follows, as the guide says (p. 82), with no data moved; JSON GENERATE and XML GENERATE place each member as they write it, and JSON PARSE as it reaches the member's pair, after any earlier pair has set a count. Refused by name: an OCCURS DEPENDING ON object that is itself variably located, and a sort key or INITIALIZE target that is (LR pp. 205, 351, 402, 454), which IBM forbids; and, keeping the message ironwork gave before, a table with items after it that is within another table or holds one, a table with variable-length elements, whose element length the interpreter keeps at its maximum",
1697 basis: Basis::Documented,
1698 oracle: Oracle::EnterpriseCobol,
1699 },
1700 Assumption {
1701 id: EXTERNAL_STORAGE,
1702 claim: "An EXTERNAL data record, and an EXTERNAL file's connector and record area, are the run unit's: the first program activated that describes one allocates it, zeroed as ironwork zeroes WORKING-STORAGE before VALUE clauses, since the record takes no VALUE (Language Reference SC27-8713-03, pp. 197, 246; Programming Guide SC27-8714-03, p. 573), and every later description shares it until the run unit ends (pp. 65, 184), in a CICS task the enclave a LINK or XCTL starts (C126). A record and a file are looked up by name in separate name spaces, and a WORKING-STORAGE record that redefines an EXTERNAL one shares its storage (p. 226). A description of another size than the one in the run unit ends the run with an ironwork abend when the program describing it is activated: the manuals say the records must define the same number of bytes, and the file descriptions the same maximum record size (pp. 186, 197), but not what the runtime does when they do not. CANCEL and an INITIAL program's return leave an EXTERNAL file open (Programming Guide pp. 178, 203); the end of the run unit closes it. INDEXED BY indexes of an EXTERNAL record stay the program's own (p. 197). Under --dialect gnucobol a later description of an EXTERNAL record shorter than the run unit's shares its storage, with a warning on standard error, as cobc's does; a longer one still ends the run",
1703 basis: Basis::Chosen,
1704 oracle: Oracle::EnterpriseCobol,
1705 },
1706 Assumption {
1707 id: GLOBAL_NAMES,
1708 claim: "A contained program sees the GLOBAL records and files of each program containing it, at that program's storage, record area and file connector (Language Reference SC27-8713-03, pp. 63-66, 185, 197). A name resolves among the program's own names and every GLOBAL name of the programs containing it, qualified as usual; where more than one item qualifies, the one declared nearest wins, the program's own first (p. 66). A file-name the program or a nearer program declares hides a farther GLOBAL one. A contained program called while a program containing it is not running, which only ironwork's flat program library allows, ends the run when it uses that program's GLOBAL names. Refused as not supported yet: a GLOBAL file whose FILE STATUS, RECORD KEY, ALTERNATE RECORD KEY or RELATIVE KEY is not a GLOBAL name of the program declaring it, as IBM resolves them there; LINAGE or REPORT on a GLOBAL file in a program that contains others, and on an EXTERNAL file; INDEXED BY in a GLOBAL record of a program that contains others, whose index is global too (p. 65); a GLOBAL file a contained program and its declaring program use differently as a print file, since ironwork decides that from each program's own WRITE statements; and SET ADDRESS OF a GLOBAL LINKAGE record from a contained program",
1709 basis: Basis::Chosen,
1710 oracle: Oracle::EnterpriseCobol,
1711 },
1712 Assumption {
1713 id: SET_TO_ENTRY,
1714 claim: "SET procedure-pointer or function-pointer TO ENTRY literal or identifier (Language Reference SC27-8713-03, 'Format 6: SET for procedure-pointer and function-pointer data items', pp. 449-450) resolves the entry when the SET runs: the program holding the name is found and loaded as a CALL of the name would find it, static for a literal under NODYNAM and dynamic for an identifier or under DYNAM (Programming Guide SC27-8714-03, pp. 557-558), and a name no program or Language Environment service has abends S806 at the SET. The manuals do not say when a dynamic entry is loaded; ironwork loads it at the SET because the pointer then holds an entry address. The pointer holds a value of ironwork's own, the same for each SET of the same name and resolution, so two pointers set to one entry compare equal; a CALL through it enters the entry as a CALL of the name does, with the arguments passed as the BY phrases say (pp. 320, 559), and ON EXCEPTION never runs, the entry having been found. After a CANCEL of the program the pointer is undefined (p. 558); ironwork's CALL through it loads the program again. identifier-9, a user-defined function returning a pointer, and SET TO a pointer a non-COBOL program set are not run, ironwork having neither",
1715 basis: Basis::Chosen,
1716 oracle: Oracle::EnterpriseCobol,
1717 },
1718 Assumption {
1719 id: HEX_CURRENCY_SIGN,
1720 claim: "CURRENCY SIGN IS X'...' gives the currency sign value the literal's bytes are in the program's code page, the CODEPAGE option's, as the Programming Guide's table of the euro sign's code points by code page has it, X'9F' in 1140 and X'5A' in 1142 (SC27-8714-03, pp. 64-65). Without PICTURE SYMBOL the literal is one byte, and the character it is must be one a PICTURE currency symbol can be; with it, the characters must include no digit, +, -, . or , (Language Reference SC27-8713-03, pp. 129-130). The symbol a PICTURE writes is the character the byte is, so the source spells it as the code page shows it; a lowercase letter, the same byte in every code page ironwork carries, keeps its case in a PICTURE as an alphanumeric symbol's does. The guide's own example, X'9F' WITH PICTURE SYMBOL 'U' (p. 64), is refused, since the Language Reference excludes U from literal-7 (p. 130)",
1721 basis: Basis::Documented,
1722 oracle: Oracle::EnterpriseCobol,
1723 },
1724 Assumption {
1725 id: INITIALIZE_FLOAT_NUMERIC,
1726 claim: "INITIALIZE's VALUE and REPLACING phrases take a COMP-1 or COMP-2 receiver as of category NUMERIC, reading the Language Reference's 'a data item of category internal floating-point ... is treated as if it were in the NUMERIC category' (SC27-8713-03, p. 352) as applying to receivers as well as to identifier-2, since rule 2 (p. 353) names no implied sending item for a floating-point category otherwise. A POINTER item, which no category names, is set to NULL with no phrase or with DEFAULT, as ironwork's INITIALIZE did before the phrases, and is left alone otherwise. The VALUE phrase finds no VALUE clause on a FILE SECTION or LINKAGE SECTION item, whose VALUE ironwork does not apply at all, and gives an OCCURS item's one VALUE to every occurrence. DBCS, EGCS, NATIONAL-EDITED and UTF-8 name no item, as ironwork has no items of those categories",
1727 basis: Basis::Chosen,
1728 oracle: Oracle::EnterpriseCobol,
1729 },
1730 Assumption {
1731 id: FLOAT_VALUE_LITERAL,
1732 claim: "A floating-point literal in the VALUE clause of a COMP-1 or COMP-2 item (Language Reference SC27-8713-03, pp. 45, 246) gives the item the value its mantissa times ten to its exponent has, written in fixed point and converted to hexadecimal floating point as MOVE converts a fixed-point literal; Enterprise COBOL converts the literal when it compiles, and its conversion may round the last hexadecimal digit differently. One whose value needs more than 31 digits in fixed point, such as 1.0E+40 or 1.0E-35, is refused, as is a floating-point literal in the VALUE of a fixed-point item, as IBM refuses it. A floating-point literal is read as one only in an item's VALUE clause, not in a level-88 VALUE or the PROCEDURE DIVISION, and a fixed-point VALUE on a COMP-1 or COMP-2 item, which IBM refuses unless it is zero, is still accepted",
1733 basis: Basis::Chosen,
1734 oracle: Oracle::EnterpriseCobol,
1735 },
1736 Assumption {
1737 id: CICS_ABEND_EXITS,
1738 claim: "HANDLE ABEND gives each logical level one exit: PROGRAM or LABEL replaces the level's exit, active, CANCEL (the default) deactivates it and RESET reactivates it, also after CICS deactivated it on entry; PUSH HANDLE suspends it until POP HANDLE (CICS TS 5.3 Application Programming Reference SC34-7402-00, pp. 314-315; Application Programming Guide SC34-7401-00, pp. 364-365, 371-373). When the task abends, the exit of the level the abend happened at, else of each level above in turn, takes the first active one, deactivated as it is entered. A LABEL is a GO TO in the program that set it, the program levels below gone (C236). A PROGRAM must be one a LINK could find when HANDLE ABEND runs, else PGMIDERR, and is entered as by LINK with the COMMAREA and EIBCALEN of the program that set the exit; when it returns, the level that set it ends and control passes to the level above, or the task ends normally, and an abend in it goes on to the levels above; one that cannot be loaded then abends APCT, which goes on likewise. Intercepted are the transaction abends: a condition's AEIx, ABEND ABCODE, ASRA for a program check and APCT; not ABEND CANCEL, ASPx or APSJ, and, by ironwork's choice, not its own IRONWORK refusals or the batch codes a CICS task does not end with here (S806 from CALL, file status abends). ASSIGN ABCODE gives the code of the abend an exit was given, spaces before any. A CALLed subprogram is at its caller's logical level (C237, C238), where RETURN and XCTL end the whole level (C233), and the program XCTL starts is at the level of the one that issued it, with the level's exit (C239)",
1739 basis: Basis::Chosen,
1740 oracle: Oracle::EnterpriseCobol,
1741 },
1742 Assumption {
1743 id: NUMCHECK_LAX_REDEFINES,
1744 claim: "Under NUMCHECK(ZON(LAX)) the two redefinitions the Programming Guide lists are recognised from the data description alone (SC27-8714-03, pp. 390-391). An unsigned zoned item, not in a table, whose level-01 or level-77 record REDEFINES a level-01 or level-77 signed zoned item whose sign is a trailing overpunch, and whose last byte is that item's last byte, is tested as a signed item is. A zoned item, unsigned or signed with a trailing overpunch and not in a table, that starts at the first byte of a level-01 or level-77 numeric-edited item its record REDEFINES may hold a space in each leading byte over a position from the edited PICTURE's start through its last leading Z, the insertion characters among those Z symbols included: ZZ,ZZ9.99 gives five, ZZ99.99 two and $ZZ9 none. Spaces and digits may mix there, as the Guide says. Insertion characters after the last leading Z, as in ZZ,999, are not tolerated, though zero suppression may leave them as spaces. The Guide shows explicit REDEFINES only, so the records of an FD, which share storage without one, are not taken as redefinitions",
1745 basis: Basis::Chosen,
1746 oracle: Oracle::EnterpriseCobol,
1747 },
1748 Assumption {
1749 id: NUMCHECK_ALWAYS_FAILS,
1750 claim: "When the compiler can determine that a NUMCHECK test will always find invalid data, it gives an error-level message and removes the test, under MSG and ABD alike (Programming Guide SC27-8714-03, pp. 388, 391); the Guide does not say when it can. ironwork determines it for an item, not in a table, that holds for the whole run what one alphanumeric VALUE clause gives it, the item's own or a group's above it, with no other VALUE clause reaching its bytes: a WORKING-STORAGE or LOCAL-STORAGE item that the INITCHECK analysis (C224) finds no statement sets and no CALL, INVOKE or EXEC statement reaches by its address, and that no table SORT or READ of a RECORD VARYING DEPENDING ON file sets. The references are those the interpreter tests that the compiler can name: an arithmetic expression's operands, PERFORM TIMES counts and expressions in a relation among them; a MOVE's sender, alphanumeric ones as integers for a numeric receiver; a numeric item compared with a number, ZERO, an expression or another numeric item, except under INVDATA(NOFORCENUMCMP); a condition-name's conditional variable; and a BY CONTENT or BY VALUE argument; each without subscripts or reference modification. The message, at the reference, names the item, its bytes and the fault, severity E (return code 8), in ironwork's words since the Guide gives no number, and the test there is removed. Any other reference keeps its test at run time, where Enterprise COBOL may have removed it. A program with object-oriented syntax is not analysed",
1751 basis: Basis::Chosen,
1752 oracle: Oracle::EnterpriseCobol,
1753 },
1754 Assumption {
1755 id: FUNCTION_SOURCE_ORDER,
1756 claim: "A run enters the first program of its source, ahead of the user-defined functions and function prototypes before it. IBM makes the first program or user-defined function of a batch compilation the default entry point, and asks for a binder ENTRY statement naming the main program when a function comes first, as IBM requires of a function with no prototype (Programming Guide, 'Structuring user-defined functions' and 'Link-editing user-defined functions'); ironwork runs as if that ENTRY statement were given. `run` and `check` compile the source's functions and prototypes with the program and report their messages first, as IBM compiles the whole compilation group; a function that does not compile stops the run, and a source of functions alone has no program to run. A function is loaded by its external name, AS literal-1 or else its function-name, matched without regard to case and without PGMNAME's truncation, as CALL matches program names",
1757 basis: Basis::Chosen,
1758 oracle: Oracle::EnterpriseCobol,
1759 },
1760 Assumption {
1761 id: FUNCTION_NAMED_AS_INTRINSIC,
1762 claim: "No user-defined function takes an intrinsic function's name, neither in FUNCTION-ID nor in the REPOSITORY paragraph without INTRINSIC. The Language Reference forbids LENGTH, RANDOM, SIGN, SUM and WHEN-COMPILED as a REPOSITORY paragraph's user-defined function names ('REPOSITORY paragraph') and says nothing of the other intrinsic names, nor whether FUNCTION name then invokes the intrinsic function or the user's; ironwork refuses the name rather than guess which, so Enterprise COBOL may compile a program ironwork refuses",
1763 basis: Basis::Chosen,
1764 oracle: Oracle::EnterpriseCobol,
1765 },
1766 Assumption {
1767 id: FUNCTION_ARGUMENT_TEMPORARIES,
1768 claim: "Conformance is between a formal parameter and a data item (Language Reference, 'USING phrase', conformance of parameters for user-defined functions): an argument that is a literal, an arithmetic expression or a function's value has no description to conform. It is evaluated before the function is entered and stored, as MOVE stores a value, in a temporary with the formal parameter's description, which the function addresses BY REFERENCE. IBM's own example passes literals so, docalc('add' 10 0.23) to PIC X(3), 999 and V999, giving result=010230 (Programming Guide, 'Invoking user-defined functions'). A BY VALUE argument, a data item's too, is stored the same way, which is COMPUTE's truncation for a numeric parameter and SET's copy for a pointer. A numeric parameter takes only what COMPUTE could send it, as IBM says of BY VALUE and ISO 1989 of every parameter: an alphanumeric, hexadecimal or national literal, or an index or numeric-edited item BY VALUE, is refused. A figurative constant is no function argument (Language Reference, 'Function-identifier'). A reference-modified data item is passed BY REFERENCE at its first byte without the conformance check, since its length is known only when the statement runs",
1769 basis: Basis::Chosen,
1770 oracle: Oracle::EnterpriseCobol,
1771 },
1772 Assumption {
1773 id: FUNCTION_SQL_CICS,
1774 claim: "The Programming Guide says SQL, CICS and JAVAIOP cannot be used with user-defined functions, as the function definition must come ahead of the program using it ('Structuring user-defined functions'), and does not say which compilations that reaches. ironwork refuses EXEC SQL and EXEC CICS in a function definition or prototype and in a program its source defines or prototypes a function before; a program that invokes a function defined in another source, with no prototype in its own source, is not refused",
1775 basis: Basis::Chosen,
1776 oracle: Oracle::EnterpriseCobol,
1777 },
1778 Assumption {
1779 id: FUNCTION_INVOCATION_ORDER,
1780 claim: "A user-defined function runs when the operand holding it is evaluated, a statement's operands in the order the interpreter evaluates them, left to right for DISPLAY and an arithmetic expression, so an operand after the invocation sees what the function stored through a BY REFERENCE argument and one before it does not; IBM documents neither when nor in what order a statement's function-identifiers are evaluated. Functions are always recursive (Programming Guide, 'Using user-defined functions'): every activation shares the function's WORKING-STORAGE and has LOCAL-STORAGE of its own, as a RECURSIVE program's do. STOP RUN in a function ends the run unit from the statement that invoked it, as in a called program",
1781 basis: Basis::Chosen,
1782 oracle: Oracle::EnterpriseCobol,
1783 },
1784 Assumption {
1785 id: INSPECT_NATIONAL_ITEM,
1786 claim: "INSPECT of a data item of usage NATIONAL, reference-modified or not, works in national characters (two-byte encoding units): TALLYING counts them, CHARACTERS takes one at a time, comparands and BEFORE or AFTER INITIAL delimiters match only at character boundaries, CHARACTERS BY takes a one-character national substitution field, and CONVERTING pairs the national characters of its operands by ordinal position (Language Reference SC27-8713-03, pp. 355-359). A figurative constant there is a one-character national literal (p. 355) of the value the Programming Guide gives (SC27-8714-03, p. 132), and as a substitution field it fills each occurrence of the subject field (p. 357). Both executors take the unit from the inspected item's kind; lowering keeps the literals of an INSPECT of a national item as values, which rt reads as national characters, and builds no CONVERTING table of bytes for one. ironwork holds no DBCS data (C212), national group, or national-edited or national numeric item, so an INSPECT of one does not arise",
1787 basis: Basis::Documented,
1788 oracle: Oracle::EnterpriseCobol,
1789 },
1790 Assumption {
1791 id: INSPECT_OPERAND_USAGE,
1792 claim: "Every identifier of an INSPECT but the count field has the inspected item's usage, and every literal is national when the inspected item is of usage NATIONAL and alphanumeric when it is of usage DISPLAY; a figurative constant not beginning with ALL takes either usage (Language Reference SC27-8713-03, p. 355; Programming Guide SC27-8714-03, pp. 112, 128). ironwork refuses each operand that breaks this as a severe error (S): with a national inspected item, a data item not of usage NATIONAL or an alphanumeric, hexadecimal or numeric literal; with any other, a national item or literal. The manuals give neither the message numbers nor their text: the messages are ironwork's. A numeric literal or ALL literal with an inspected item that is not national is still taken as its characters, as before this entry. A function-identifier operand is not checked when compiled, since ironwork knows an intrinsic function's category only when it runs (C232)",
1793 basis: Basis::Documented,
1794 oracle: Oracle::EnterpriseCobol,
1795 },
1796 Assumption {
1797 id: INSPECT_OPERAND_MADE_NATIONAL,
1798 claim: "Where INSPECT works in national characters, of a national item (C230) or a national function value (C191), an operand whose value is not national, an alphanumeric function value or a literal of a TALLYING of a function value, is converted to national characters through the program's code page, as MOVE converts an alphanumeric sender to a national receiver. IBM requires such an operand to be national (Language Reference SC27-8713-03, p. 355) and says nothing of a run that has one, so the conversion is ironwork's choice. A national value as an operand of an INSPECT of an item that is not national is compared byte by byte",
1799 basis: Basis::Chosen,
1800 oracle: Oracle::EnterpriseCobol,
1801 },
1802 Assumption {
1803 id: CICS_ABEND_LABEL_GO_TO,
1804 claim: "When a HANDLE ABEND LABEL exit is taken in a COBOL program, control returns to the HANDLE ABEND command with the registers restored and a GO TO is executed there (CICS TS 6.x, HANDLE ABEND, topic dfhp4_handleabend; 'Creating a program-level abend program or routine', dfhp3mh). The return points of out-of-line PERFORMs are not registers but the program's own state, which a CALL resets (Programming Guide SC27-8714-03, p. 547), so the PERFORMs the abend left keep theirs armed, as a GO TO out of them would (C99): control that later passes the end of such a range returns after its PERFORM, which puts back the return point it displaced, and a PERFORM that repeats or is inside another statement is refused there as C99 says. The GO TO is the HANDLE ABEND command's, so a debugging section on the label gets that command's line as DEBUG-LINE; the manuals say nothing of PERFORMs here",
1805 basis: Basis::Chosen,
1806 oracle: Oracle::EnterpriseCobol,
1807 },
1808 Assumption {
1809 id: CICS_ABEND_EXIT_ACROSS_CALL,
1810 claim: "A CALLed subprogram runs at its caller's logical level (CICS TS 6.x, 'Flow of control between programs and subprograms', dfhp3_cobol_subprog_flow). A static CALL, and a CALL of a contained program, leaves the level's HANDLE ABEND exit in effect in the subprogram, and an exit the subprogram sets is the level's after it returns ('Rules for calling subprograms', dfhp3_cobol_subprog_rules: for a statically called program, abend handling remains in effect irrespective of CBLPSHPOP). A dynamic CALL runs as under CBLPSHPOP(ON), the default (Enterprise COBOL 6.4 Performance Tuning Guide, CBLPSHPOP): the caller's exit is suspended, as by PUSH HANDLE, until the subprogram returns, when the exit the subprogram set is dropped and the caller's put back, as by POP HANDLE; an abend in the subprogram meets the exit as the subprogram left it. The Programming Guide describes that PUSH for a CALL of any program that is not contained and says nothing of static calls (SC27-8714-03, p. 503); ironwork follows the CICS rule. HANDLE CONDITION, IGNORE CONDITION and PUSH HANDLE's stack go with a CALL in the same way (C234)",
1811 basis: Basis::Chosen,
1812 oracle: Oracle::EnterpriseCobol,
1813 },
1814 Assumption {
1815 id: CICS_ABEND_LABEL_OWNER,
1816 claim: "A HANDLE ABEND LABEL exit is taken only in the activation of the program that set it, when an abend reaches the logical level there, from that program or from a level below it. An abend that reaches the level in another program, one the setter CALLed, one that CALLed it or one that took its place by XCTL, or after the setter has returned, abends the task APC2, the code CICS gives for an illegal branch following an abend with an active handle label abend, an out-of-block GO TO to an inactive block (CICS TS 6.x, abend code APC2); the Programming Guide says a HANDLE LABEL cannot handle an abend caused by another program invoked with CALL, and that such cross-program branching ends the transaction (SC27-8714-03, p. 503). The exit is deactivated as when taken, ASSIGN ABCODE gives the code it was reached with, and APC2 goes on to the levels above. A program CALLed again is a new activation, which does not take a label its earlier one set. A PROGRAM exit is entered where the logical level runs, whichever of its programs set the exit or abended",
1817 basis: Basis::Chosen,
1818 oracle: Oracle::EnterpriseCobol,
1819 },
1820 Assumption {
1821 id: CICS_ABEND_EXIT_ACROSS_XCTL,
1822 claim: "XCTL transfers control to another program at the same logical level and releases the program that issued it (CICS TS 6.x, XCTL, dfhp4_xctl). The HANDLE ABEND exit belongs to the logical level, not to a program: a HANDLE ABEND overrides any preceding one in any application program at the same logical level, and only one exit at each level can be active ('Abnormal termination recovery', dfhp378), where HANDLE CONDITION applies only to the program that issued it and is deactivated by LINK and XCTL (dfhp4_handlecondition, dfhp3_exc_handlecondition). The program XCTL starts therefore has the level's exit; the manuals do not say XCTL cancels it. A PROGRAM exit is entered with the COMMAREA of the program that issued the HANDLE ABEND, not of the one that abended (dfhp4_handleabend): ironwork keeps that program's DFHCOMMAREA and EIBCALEN when the command runs, and when it had none the program running the level gives its own. A LABEL exit abends APC2 when an abend reaches it (C238), the program that set it having been released",
1823 basis: Basis::Chosen,
1824 oracle: Oracle::EnterpriseCobol,
1825 },
1826 Assumption {
1827 id: CICS_RETURN_ENDS_THE_LEVEL,
1828 claim: "EXEC CICS RETURN and XCTL end the logical level they are issued at, whichever of its programs issues them. A program a COBOL CALL reaches, static or dynamic, is at its caller's logical level, and a run unit is what the task, a LINK or an XCTL starts, with the programs it CALLs (CICS TS 6.x, 'Flow of control between programs and subprograms', dfhp3_cobol_subprog_flow). RETURN in a CALLed program terminates the calling program ('Rules for calling subprograms', dfhp3_cobol_subprog_rules) and goes back to the program that LINKed to the level, or at the task's first level to CICS (dfhp3_cobol_subprog_flow; RETURN, dfhp4_return); TRANSID and COMMAREA are allowed there, where RETURN goes back to CICS, as for the program running the level; below it COMMAREA raises INVREQ and TRANSID is allowed (C143). XCTL in a CALLed program starts its program as the one running the level, the CALL chain released with the run unit (XCTL, dfhp4_xctl), with a copy of the COMMAREA, as every XCTL passes, and the exit C239 gives; the level ends when that program does. Each program of the CALL chain then ends as its CALL comes back, as by GOBACK there: nothing after the CALL runs, NOT ON EXCEPTION and the RETURNING item included, and the PERFORMs in progress end with the activations whose return points they are (C99). The manuals do not say what becomes of the CALL statements; ironwork ends their programs so",
1829 basis: Basis::Chosen,
1830 oracle: Oracle::EnterpriseCobol,
1831 },
1832 Assumption {
1833 id: CICS_HANDLERS_ACROSS_CALL,
1834 claim: "HANDLE CONDITION, IGNORE CONDITION and the PUSH HANDLE stack belong to the logical level, as the HANDLE ABEND exit does (C237): a program a CALL reaches is at its caller's level and inherits the current HANDLE commands, and POP HANDLE undoes the last PUSH HANDLE at the current link level (CICS TS 6.x, PUSH HANDLE, dfhp4_pushhandle; POP HANDLE, dfhp4_pophandle). For a static CALL, and a CALL of a contained program, condition, AID and abend handling remain in effect whatever CBLPSHPOP says ('Rules for calling subprograms', dfhp3_cobol_subprog_rules): the subprogram starts with the caller's handlers and stack, and the caller has them as the subprogram left them, its HANDLE CONDITION, IGNORE CONDITION, PUSH and POP HANDLE included, as with a CALL that pushes nothing, where the caller inherits any settings made in the subprogram (Enterprise COBOL 6.4 Performance Tuning Guide, CBLPSHPOP). A dynamic CALL runs under CBLPSHPOP(ON), the default: COBOL issues a PUSH HANDLE on entry, after which no condition or abend handling is active in the subprogram until it issues its own, and a POP HANDLE when control returns, which drops what the subprogram set and puts the caller's back (dfhp3_cobol_subprog_rules; Programming Guide SC27-8714-03, p. 503). ironwork takes those as a PUSH and a POP on the level's stack: a POP HANDLE in the subprogram with no PUSH of its own undoes the CALL's, the return's POP then undoes the PUSH before it, and with none left, which the manuals do not cover, changes nothing. A subprogram that abends does not return: the abend meets the handlers as it left them",
1835 basis: Basis::Chosen,
1836 oracle: Oracle::EnterpriseCobol,
1837 },
1838 Assumption {
1839 id: CICS_CONDITION_LABEL_OWNER,
1840 claim: "A HANDLE CONDITION label, like a HANDLE ABEND one (C238), belongs to the program activation that issued the HANDLE CONDITION. The label must be in the same PROCEDURE DIVISION as the command that causes the branch, a HANDLE label cannot handle a condition caused by another program invoked with CALL, and the attempt at cross-program branching ends the transaction (Programming Guide SC27-8714-03, pp. 503-504). A condition raised where the label it goes to was set by another activation, a program the CALL passed the handlers from or to, or one that has returned, therefore abends the task APC2, the code CICS ends a dynamically called program with when it abends under CBLPSHPOP(OFF) with its caller's condition handling active (dfhp3_cobol_subprog_rules; CICS TS 6.x, abend code APC2), and a HANDLE ABEND exit can intercept it (C24). The Programming Guide says a condition in a nested program goes to its container's label with unpredictable results (p. 504); ironwork abends APC2 there too. IGNORE CONDITION, which names no label, applies in whichever program raises the condition, and a program CALLed again is a new activation, which does not take a label its earlier one set",
1841 basis: Basis::Chosen,
1842 oracle: Oracle::EnterpriseCobol,
1843 },
1844 Assumption {
1845 id: CICS_RETURN_BELOW_THE_FIRST_LEVEL,
1846 claim: "The COMMAREA, IMMEDIATE and CHANNEL options of RETURN can be used only when RETURN returns control to CICS; otherwise INVREQ occurs with RESP2 2, 'a RETURN command with the CHANNEL, COMMAREA, or IMMEDIATE option is issued by a program that is not at the highest logical level' (CICS TS 6.x, RETURN, dfhp4_return). TRANSID is not among them: specified on a program that is not at the highest level, it is the transaction identifier for the terminal's next input unless an error on COMMAREA, INPUTMSG or CHANNEL on the final RETURN clears it (dfhp4_return). A RETURN TRANSID below the first level therefore names the next transaction, which a later RETURN TRANSID replaces, and one that raises INVREQ names none. ironwork raises none of the errors that clear it, models no terminal-less task (RESP2 1) and keeps CHANNEL only for the INVREQ",
1847 basis: Basis::Documented,
1848 oracle: Oracle::EnterpriseCobol,
1849 },
1850 Assumption {
1851 id: CICS_STOP_RUN_ENDS_THE_LEVEL,
1852 claim: "STOP RUN terminates the run unit, the Language Environment enclave, whose main routine it ends (z/OS 3.1 Language Environment Programming Guide, 'The enclave defines the scope of language statements', ceea200118). Under CICS a run unit is what the task, a LINK or an XCTL starts, with the programs it CALLs ('CICS run unit', ceea200254; CICS TS 6.x, 'Flow of control between programs and subprograms', dfhp3_cobol_subprog_flow), and a program at level 2, LINKed, CALLed there or started there by XCTL, can use GOBACK, STOP RUN or EXEC CICS RETURN to return to the level 1 program that LINKed to it (dfhp3_cobol_subprog_flow). STOP RUN therefore ends the logical level it runs at, as RETURN does (C233): a LINK, or a HANDLE ABEND PROGRAM exit entered as by LINK, comes back from it, and at the task's first level it ends the task. The Programming Guide's table of termination statements says instead that in a CICS environment STOP RUN terminates the entire transaction, including all programs running within it (SC27-8714-03, p. 546, Table 70); ironwork follows CICS and Language Environment, whose rules for the level are the more specific",
1853 basis: Basis::Chosen,
1854 oracle: Oracle::EnterpriseCobol,
1855 },
1856 Assumption {
1857 id: CICS_RUN_UNIT_PER_LINK,
1858 claim: "Under CICS a run unit, the Language Environment enclave, is entered at the start of the task or by a LINK or XCTL, and each enclave has its own heap storage and other Language Environment resources (CICS TS 6.x, 'Flow of control between programs and subprograms', dfhp3_cobol_subprog_flow; z/OS 3.1 Language Environment Programming Guide, 'CICS run unit', ceea200254). On each entry to a LINKed program a new initialized copy of its WORKING-STORAGE is provided and the run unit is reinitialized; a program a static or dynamic CALL reaches gets a new initialized copy on its first entry within a logical level and its last-used state on later entries at the same level ('Rules for calling subprograms', dfhp3_cobol_subprog_rules). ironwork therefore keeps each loaded program's state, its storage, open files, CLOSE WITH LOCK, ALTERs, activity and whether a dynamic CALL entered it, per CICS run unit: a LINK, an XCTL and a HANDLE ABEND PROGRAM exit, entered as by LINK, each start one where every program starts in its initial state with storage of its own, and when it ends the files its programs left open are closed, as an enclave's are, and the run unit that started it has its programs as it left them. A program active in a higher run unit can be CALLed in a lower one. CANCEL and INITIAL act within the run unit. XCTL passes control to a new enclave at the same logical level (ceea200254), so a program CALLed both before and after an XCTL starts afresh after it, which the logical-level wording of the rules leaves open. The files are closed as when STOP RUN ends an enclave (ceea200118). EXTERNAL data and Language Environment heap storage belong to the enclave too (C126); the objects of object-oriented programs stay with the task",
1859 basis: Basis::Documented,
1860 oracle: Oracle::EnterpriseCobol,
1861 },
1862 Assumption {
1863 id: CICS_HANDLERS_ACROSS_XCTL,
1864 claim: "HANDLE CONDITION and IGNORE CONDITION apply only to the program in which they are specified and remain active while it is being executed (CICS TS 6.x, HANDLE CONDITION, dfhp4_handlecondition; IGNORE CONDITION, dfhp4_ignorecondition), and when control passes to another program by LINK or XCTL the HANDLE CONDITION commands active in the calling program are deactivated ('Using the HANDLE CONDITION command', dfhp3_exc_handlecondition). XCTL releases the program that issues it (XCTL, dfhp4_xctl), so the program it starts has none of them. The HANDLE ABEND exit belongs to the logical level and stays (C239). POP HANDLE restores the state before a PUSH HANDLE executed at the current link level, and raises INVREQ only when no such PUSH has been executed (POP HANDLE, dfhp4_pophandle); XCTL keeps the link level, and the manuals do not say it discards the stack. ironwork keeps the level's PUSH HANDLE stack across XCTL, with the HANDLE ABEND exit each entry suspended, and empties each entry's HANDLE CONDITION and IGNORE CONDITION, which belonged to the released program: a POP HANDLE in the new program undoes a PUSH the released one made and restores that exit, with no condition handling. HANDLE AID, which ironwork accepts and does not act on, has no state to carry",
1865 basis: Basis::Chosen,
1866 oracle: Oracle::EnterpriseCobol,
1867 },
1868 Assumption {
1869 id: CICS_NO_OBJECT_ORIENTED_COBOL,
1870 claim: "COBOL programs that have object-oriented syntax for Java interoperability cannot run in CICS (Programming Guide SC27-8714-03, p. 495), and COBOL class definitions and methods cannot contain EXEC CICS statements, cannot be run in CICS and cannot be compiled with the CICS option (p. 679); the CICS reserved-word table flags INVOKE, METHOD, OBJECT and FACTORY (p. 502). An INVOKEd method therefore shares no CICS handlers with the program that invokes it, and starts with none either: it does not run. ironwork compiles such a program as before and refuses an INVOKE that a CICS task reaches, before any operand is evaluated, with an IRONWORK abend",
1871 basis: Basis::Documented,
1872 oracle: Oracle::EnterpriseCobol,
1873 },
1874 Assumption {
1875 id: CICS_TRANSFER_TO_A_RUNNING_PROGRAM,
1876 claim: "XCTL releases the program that issues it, and the program it transfers control to is loaded only if it is not already in main storage (CICS TS 6.x, XCTL, dfhp4_xctl); none of XCTL's conditions is for a target that is running or in the CALL chain. XCTL and LINK each pass control to a new Language Environment enclave (z/OS 3.1 Language Environment Programming Guide, 'CICS run unit', ceea200254), where the target starts afresh (C145). So XCTL to the program that issues it, to one in the CALL chain it releases, or to the task's first program, runs that program from its start with new storage, and LINK to a program running at a higher level, the task's first program included, runs a copy of its own; a HANDLE ABEND PROGRAM exit naming one is entered the same way. ironwork's run unit holds no handle for the task's first program, so each activation carries a reference to it for these transfers and for a CALL of it (C127)",
1877 basis: Basis::Documented,
1878 oracle: Oracle::EnterpriseCobol,
1879 },
1880 Assumption {
1881 id: CICS_ENCLAVE_EXTERNALS_AND_HEAP,
1882 claim: "External data persists over the lifetime of an enclave, and its scope is the enclosing enclave, COBOL EXTERNAL data items included (z/OS 3.1 Language Environment Programming Guide, 'The enclave defines the scope and visibility of the following types of data', ceea200117; Language Environment, 'Name scope of external data', cee142377: under COBOL the enclave). Each enclave has its own heap storage, and a LINK or XCTL passes control to a new enclave ('CICS run unit', ceea200254); the storage Language Environment uses for its heap at each CICS link level is freed when the program terminates (CICS TS 6.x, 'Language Environment storage', dfhp3_langenv_storage), and Language Environment generally does not allow language file sharing across enclaves (dfhp3_langenv_oview). A LINK, an XCTL and a HANDLE ABEND PROGRAM exit therefore start a run unit (C145) with no EXTERNAL record or file and an empty heap: the first program there that describes an EXTERNAL record or file gets it afresh, zeroed (C180), and CEEGTST takes storage from the new heap. When the run unit ends, the EXTERNAL files it left open are closed with its programs' files, its EXTERNAL data and heap blocks are gone, and the run unit above has its own back. The manuals do not say what CEEFRST does with a block another enclave got; ironwork gives CEE0810, the condition for an address that is not heap storage, both for a block of the run unit above and for one of a run unit that has ended. A batch run is one enclave and is unchanged",
1883 basis: Basis::Documented,
1884 oracle: Oracle::EnterpriseCobol,
1885 },
1886 Assumption {
1887 id: RECURSIVE_CALL_OF_AN_ACTIVE_PROGRAM,
1888 claim: "With the RECURSIVE clause a program can be reentered while a previous invocation is still active, and without it an active program cannot be (Language Reference SC27-8713-03, p. 102); only a RECURSIVE program can execute a CALL that directly or indirectly calls itself (p. 318). A recursive CALL of a program without RECURSIVE signals a condition, and if it is unhandled the run unit ends (Programming Guide SC27-8714-03, p. 557): IGZ0064S, 'A recursive call to active program program-name in compilation unit compilation-unit was attempted', and the application is terminated (z/OS 3.1 Language Environment Runtime Messages, IGZ0064S, cs00499). ironwork ends the run with U4038 and that message, as it ends one for an SSRANGE condition nothing handles (L19), naming the outermost program of the CALLed program's source as the compilation unit. The run unit's first program is no exception: a RECURSIVE main program can CALL itself, as the Programming Guide's factorial program does (p. 15), finding its WORKING-STORAGE in its last-used state, and a program stays active when a CALL of it returns while an earlier activation of it is still running. Under CICS a program is active only in its own run unit (C145): a CALL of the task's first program in a run unit a LINK or XCTL started runs a fresh copy, and one in the first program's own run unit is a recursive call. ironwork's run unit holds no handle for the first program, so each activation carries a reference to it (C148); a function or a method, whose activation does not, cannot CALL it, an ironwork refusal",
1889 basis: Basis::Documented,
1890 oracle: Oracle::EnterpriseCobol,
1891 },
1892 Assumption {
1893 id: CICS_RETURN_COMMAREA_LENGTH,
1894 claim: "The valid range for the length of the COMMAREA RETURN passes is 0 through 32763 bytes, and outside it the LENGERR condition occurs, RESP2 11, whose default action terminates the task abnormally (CICS TS 6.x, RETURN, dfhp4_return), with AEIV. The length is LENGTH when it is given, and otherwise the COMMAREA item's, as the translator supplies it. A TRANSID specified below the highest level is cleared when there is an error on COMMAREA on the final RETURN (dfhp4_return); ironwork clears the next TRANSID on that LENGERR whichever program named it, the failing RETURN included, and a RETURN that RESP or a handler goes on from has set neither TRANSID nor COMMAREA. COMMAREA below the highest level raises INVREQ with RESP2 2 whatever its length (C143); the manual does not order the two conditions, and ironwork tests the level first, so LENGERR comes only from the RETURN to CICS. A LENGTH greater than the COMMAREA item, which the manual says gives unpredictable results and may give LENGERR, passes the item's bytes. A LINKAGE item with no address raises RESP2 26, a COMMAREA address of zero with a length that is not, as it does for LINK and XCTL (C103), and with LENGTH(0) passes a COMMAREA of no bytes",
1895 basis: Basis::Documented,
1896 oracle: Oracle::EnterpriseCobol,
1897 },
1898 Assumption {
1899 id: CICS_RUN_UNIT_STORAGE_RELEASED,
1900 claim: "Running in CICS, a reentrant COBOL program's WORKING-STORAGE is allocated from heap storage (Programming Guide SC27-8714-03, p. 40) and persists until the end of the run unit or until the program is cancelled (p. 479); each enclave has its own heap (z/OS 3.1 Language Environment Programming Guide, 'CICS run unit', ceea200254; C126), freed when the program at the CICS link level terminates (CICS TS 6.x, 'Language Environment storage', dfhp3_langenv_storage). When the run unit a LINK, an XCTL or a HANDLE ABEND PROGRAM exit started ends (C145), ironwork therefore releases the storage of every program activated in it with the rest of the memory the run unit took. A program first loaded there stays loaded, as CICS keeps a program in main storage once loaded (XCTL, dfhp4_xctl), with no storage of its own: its next activation, in a later LINK's run unit or by a CALL at a higher level, gets new storage in its initial state. A pointer the level above kept to the released storage no longer reaches the program's, which on z/OS has been freed. The program a LINK or XCTL names is loaded before its run unit starts, in the run unit that issued the command, and keeps the storage loading gave it there",
1901 basis: Basis::Documented,
1902 oracle: Oracle::EnterpriseCobol,
1903 },
1904 Assumption {
1905 id: TRAP_OFF_LEAVES_FILES_OPEN,
1906 claim: "When the run unit ends normally, all open files are closed, and when it ends abnormally they are closed if TRAP(ON) is in effect (Programming Guide SC27-8714-03, p. 203), Language Environment's default (z/OS 3.1 Language Environment Programming Reference, TRAP, ceea300/cltrap). Under TRAP(OFF), Language Environment is not told of a program check or an abend and does not close the files high-level languages opened, so records might be lost (cltrap), and the VSAM CLOSE the abend invokes does not update the data set's catalog information (z/OS 3.1 DFSMS Using Data Sets, idad400/clds9), which leaves a VSAM data set opened for output marked open (C220). ironwork reads TRAP from the runtime options of a job step's PARM or run's --parm, after the last slash (C250), the last TRAP there deciding, as cltrap has the last of several STAE or SPIE decide; a CICS task takes the default. ironwork takes S0C4 to S0CF and S322 as the program checks and abends TRAP(OFF) hides; a failing I/O status or a U code ends the run through a condition Language Environment handles under either setting (CEESGL is unaffected by TRAP, cltrap), closing the files as ironwork's own stops do. A file left open is written as CLOSE writes it, every record reaching the data set, where on z/OS buffers are not flushed",
1907 basis: Basis::Chosen,
1908 oracle: Oracle::EnterpriseCobol,
1909 },
1910 Assumption {
1911 id: CICS_TRANSFER_COMMAREA_LENGTH,
1912 claim: "LINK and XCTL raise LENGERR, whose default action terminates the task abnormally, with AEIV, with RESP2 11 for a COMMAREA length less than 0 or greater than the permitted length, and RESP2 26 for a COMMAREA address of zero with a length that is not (CICS TS 6.x, LINK, dfhp4_link; XCTL, dfhp4_xctl). XCTL's limit is 32763; LINK says only 'the permitted length', and ironwork takes 32763 for it too, the limit XCTL and RETURN state (dfhp4_return, C128) and the longest COMMAREA the EXCI LINK allows (dfhtm4w). The length is LENGTH, else the COMMAREA item's, which the translator moves to the length argument and passes with the item BY REFERENCE ('Translated code for EXEC CICS commands', dfhp4_translatedcode). A LINKAGE item with no address, such as the DFHCOMMAREA of a program given no COMMAREA, whose parameter list entry is X'00000000' (dfhp4_translatedcode), therefore reaches CICS at address zero: it raises RESP2 26, not the ASRA a reference to it gives, and with LENGTH(0) passes no COMMAREA. With no COMMAREA option, EIBCALEN is zero ('Passing data to other programs by using COMMAREA', dfhp37u) and LENGTH is not read. The manuals do not order LENGERR and PGMIDERR; ironwork tests the COMMAREA first. A program the library does not hold raises PGMIDERR with RESP2 1, a program with no installed resource definition and no autoinstall, as HANDLE ABEND PROGRAM does (dfhp4_handleabend): ironwork has no program definitions. Not raised: RESP2 12 and 13, for DATALENGTH, which CICS checks only for a remote or dynamic LINK (dfhp4_link); 27 and 28, for INPUTMSGLEN and for a destructive overlap while copying; INVREQ's RESP2 values, for INPUTMSG, SYSID, SYNCONRETURN, TRANSID, channels, Java, applications on platforms, GLUEs and TRUEs, PLT programs and DPL; NOTAUTH; and PGMIDERR 2, 3, 9 and 21 to 27, for disabled, unloadable and remote program definitions, autoinstall, dynamic routing and Liberty, none of which ironwork has",
1913 basis: Basis::Documented,
1914 oracle: Oracle::EnterpriseCobol,
1915 },
1916 Assumption {
1917 id: CICS_RANDOM_PER_RUN_UNIT,
1918 claim: "If the first reference to FUNCTION RANDOM in the run unit does not specify argument-1, the seed value used is zero, an argument starts a new sequence, and later references without one return the next number in the current sequence (Language Reference SC27-8713-03, 'RANDOM', p. 621). Under CICS a LINK, an XCTL and a HANDLE ABEND PROGRAM exit each start a run unit, a Language Environment enclave of its own (C145), so the sequence is the run unit's: the first reference without an argument there starts from seed zero whatever the run unit that issued the command has drawn, a seed given there starts that run unit's sequence only, and when it ends the issuing run unit goes on with its own sequence where it left it. A batch run is one run unit and is unchanged",
1919 basis: Basis::Documented,
1920 oracle: Oracle::EnterpriseCobol,
1921 },
1922 Assumption {
1923 id: CICS_RETURN_CODE_PER_RUN_UNIT,
1924 claim: "When a COBOL program ends, its RETURN-CODE goes to the operating system, or to the calling program, whose own RETURN-CODE is set to it; RETURN-CODE is implicitly a binary halfword with VALUE ZERO (Language Reference SC27-8713-03, p. 24). A LINK is not a CALL: when the enclave a LINK or XCTL started ends without the CICS thread ending, its return code is placed in the RESP2 field of the LINK or XCTL and its reason code is discarded (z/OS 3.1 Language Environment Programming Guide, 'Finding the return and reason code from the enclave', ceea200407), the return code being the user return code, a COBOL program's RETURN-CODE, plus the reason code ('How the Language Environment enclave return code is calculated', ceea200105), which is zero for a normal end. Each run unit a LINK, an XCTL or a HANDLE ABEND PROGRAM exit starts (C145) therefore begins with RETURN-CODE zero, and a LINK that comes back sets EIBRESP2 and RESP2 to the RETURN-CODE its run unit ended with, by RETURN or by STOP RUN (C144), with EIBRESP and RESP zero again whatever its programs' commands left, and leaves the issuing program's RETURN-CODE as it was. XCTL releases the program that issues it (dfhp4_xctl), so its RESP2 is never read; ironwork gives the LINK above the level, or the end of the task, the RETURN-CODE the program XCTL started ends with. The manuals do not say what the CALL the translator makes of a command does to the issuing program's RETURN-CODE, and ironwork leaves it as it was for every command. A batch run is one run unit and is unchanged, its programs sharing one RETURN-CODE",
1925 basis: Basis::Documented,
1926 oracle: Oracle::EnterpriseCobol,
1927 },
1928 Assumption {
1929 id: CICS_ENTRY_POINTERS_ACROSS_RUN_UNITS,
1930 claim: "A function-pointer or procedure-pointer SET TO ENTRY holds a value of ironwork's own naming the entry (C140), and the task keeps one list of those entries rather than one per CICS run unit: a pointer set in one run unit and passed to another, in a COMMAREA or by a CALL, enters the entry as a CALL of its name in the run unit that CALLs through it does, where the program starts afresh if it has not run there (C145), and so does one set in a run unit that has since ended. On z/OS the pointer holds the entry's address, and CICS keeps a program in main storage once it is loaded (XCTL, dfhp4_xctl). The manuals say only that a pointer to an entry of a program later cancelled is undefined (Programming Guide SC27-8714-03, p. 558), not what a pointer means in another enclave",
1931 basis: Basis::Chosen,
1932 oracle: Oracle::EnterpriseCobol,
1933 },
1934 Assumption {
1935 id: INITIALIZE_REFERENCE_MODIFIED,
1936 claim: "INITIALIZE of a reference-modified identifier has one receiver, the unique data item reference modification creates, which is an elementary item of category alphanumeric, national for a national item and alphabetic for an alphabetic one (Language Reference SC27-8713-03, pp. 75-76). Only its character positions change: a group is not walked for its elementary items, and a numeric DISPLAY item takes SPACE there, the implied sending item for those categories, not ZERO (pp. 351-353). The unique data item has no data description entry, so the VALUE phrase finds no VALUE clause for it (p. 352): the VALUE phrase alone leaves it unchanged, REPLACING moves to it when it names its category, and DEFAULT or no phrase gives it SPACE. The compiler's IGYPS2047-W check takes the same category",
1937 basis: Basis::Documented,
1938 oracle: Oracle::EnterpriseCobol,
1939 },
1940 Assumption {
1941 id: SORT_INVALID_DIGIT_ABENDS,
1942 claim: "A SORT step that edits or converts a ZD or PD field holding a digit nibble above 9 ends with a data exception, S0C7. DFSORT's manual gives two outcomes for an invalid digit, 'a data exception (0C7 ABEND) or incorrect numeric output' (z/OS 3.1 DFSORT Application Programming Guide, OUTFIL OUTREC, p,m,f,edit, Table 7 notes), without saying which happens when; ironwork takes the abend",
1943 basis: Basis::Chosen,
1944 oracle: Oracle::EnterpriseCobol,
1945 },
1946 Assumption {
1947 id: SORT_IFTHEN_FIXED_LENGTH,
1948 claim: "Fixed-length records an INREC, OUTREC or OUTFIL IFTHEN edit makes all take one length: IFOUTLEN when it is given, else the longest of the input record and of the record each clause's BUILD, OVERLAY or PUSH items make, the shorter records padded with blanks. DFSORT says it 'sets an appropriate LRECL ... based on the build, overlay, find/replace and group operation items specified by the IFTHEN clauses' and 'does not analyze the possible results of WHEN=(logexp) conditions' (INREC control statement, IFTHEN and IFOUTLEN), without giving the rule; this one is chosen to match it",
1949 basis: Basis::Chosen,
1950 oracle: Oracle::EnterpriseCobol,
1951 },
1952 Assumption {
1953 id: SORT_MASK_GROUPS_OF_THREE,
1954 claim: "The edit masks M0-M5 and M10-M26 are built as 31-digit patterns whose integer digits are grouped in threes from the decimal point or the right, as Table 8 of OUTFIL OUTREC shows them and as Table 11's output lengths imply. Table 8 prints M22 as SI III III III III IIII III III III IIT,TT, with one group of four; ironwork takes that group as three, the M22 length d + 1 + d/3 holding only then",
1955 basis: Basis::Chosen,
1956 oracle: Oracle::EnterpriseCobol,
1957 },
1958 Assumption {
1959 id: SORT_PATTERN_DECIMAL_POINT,
1960 claim: "In an EDIT or EDxy pattern, a period immediately followed by a digit position is the significant decimal point DFSORT names: digits before the first nonzero insignificant digit, significant digit or significant decimal point become blanks, and an insignificant digit after one is shown (OUTFIL OUTREC, edit patterns). The manual does not define which period is significant; so EDIT=(III.II) edits 5 as .05 and EDIT=(IIT.TT) as 0.05",
1961 basis: Basis::Chosen,
1962 oracle: Oracle::EnterpriseCobol,
1963 },
1964 Assumption {
1965 id: ALTERNATE_KEYS_FROM_THE_BASE,
1966 claim: "A program's ALTERNATE RECORD KEY reads and writes go to the base cluster's records as they stand, as an alternate index in the upgrade set holds them. Each alternate index path has its own DD, named from the base cluster's ddname with 1, 2 and on (Programming Guide SC27-8714-03, 'Allocating VSAM files'). ironwork does not open those DDs. An alternate index defined NOUPGRADE or never built gives the program the same records as a current one, and a key that does not match the catalog, which gives OPEN status 39 on z/OS ('Opening a file'), is not checked",
1967 basis: Basis::Chosen,
1968 oracle: Oracle::EnterpriseCobol,
1969 },
1970 Assumption {
1971 id: PATH_PRESENTS_THE_BASE,
1972 claim: "A DD that names a path gives the program the base cluster's records in the order of the path's alternate index: by alternate key, and under one key in the order its prime keys are stored. A path over the cluster itself gives them in prime key order. A prime key the base cluster no longer holds reaches no record. What the program writes through the path goes back to the base cluster by prime key when the step ends. The path's alternate index is then rebuilt, and so is the rest of the upgrade set when the path is defined with UPDATE",
1973 basis: Basis::Chosen,
1974 oracle: Oracle::EnterpriseCobol,
1975 },
1976 Assumption {
1977 id: GENERATED_COMPONENT_NAMES,
1978 claim: "A data or index component that DEFINE does not name gets the name VSAM generates (z/OS 3.1 DFSMS Using Data Sets, 'Naming a cluster'): a last qualifier CLUSTER becomes DATA or INDEX, a name of up to 38 characters gains .DATA or .INDEX, and one of 39 to 42 gains .D or .I. A longer name keeps its first four qualifiers at most and gains qualifiers VSAM makes up. ironwork derives those from the name, so they differ from a z/OS catalog's",
1979 basis: Basis::Chosen,
1980 oracle: Oracle::EnterpriseCobol,
1981 },
1982 Assumption {
1983 id: BLDINDEX_NON_ENDING_ERRORS,
1984 claim: "BLDINDEX's non-ending errors end the command with condition code 4: a base record too short for the alternate key (IDC1644I), a repeated key in a UNIQUEKEY alternate index (IDC1645I, once for each prime key it leaves out), and prime keys past the record size (IDC1646I), each followed by IDC1653I. IBM's message pages give no code, and in the condition-code table 4 is a function that continued after a warning. An alternate index record starts with a five-byte header: X'01' for prime-key pointers, a halfword count, the pointer length and the key length. IBM gives the header's length and contents but not their order",
1985 basis: Basis::Chosen,
1986 oracle: Oracle::EnterpriseCobol,
1987 },
1988 Assumption {
1989 id: LISTCAT_WHAT_THE_CATALOG_KEEPS,
1990 claim: "LISTCAT names the catalog IRONWORK.CATALOG and lists every data set in the data set directory that is not a VSAM object or a generation data group as NONVSAM. ALL lists what ironwork's catalog keeps: associations, KEYLEN, RKP, AXRKP, AVGLRECL, MAXLRECL, the organization, key uniqueness, REC-TOTAL, UPGRADE or UPDATE, and a generation data group's LIMIT, SCRATCH and EMPTY. It leaves out HISTORY, SMSDATA, RLSDATA, PROTECTION, space, volumes and control interval sizes, which ironwork does not keep",
1991 basis: Basis::Chosen,
1992 oracle: Oracle::EnterpriseCobol,
1993 },
1994 Assumption {
1995 id: UPGRADE_SET_AFTER_THE_STEP,
1996 claim: "VSAM brings an alternate index in the upgrade set up to date as each record is written. ironwork rebuilds it from the base cluster when a step that changed the base cluster's data set ends, abended or not, and after a REPRO into the base cluster. A program sees no difference between the two, since its own ALTERNATE RECORD KEY reads use the base cluster (C350)",
1997 basis: Basis::Chosen,
1998 oracle: Oracle::EnterpriseCobol,
1999 },
2000 Assumption {
2001 id: BLDINDEX_REFUSALS,
2002 claim: "BLDINDEX needs a base cluster with at least one record, and an empty alternate index or one defined with REUSE (z/OS 3.1 DFSMS Access Method Services, BLDINDEX). Otherwise ironwork refuses it with its own message and condition code 12. IBM documents both conditions but not the message or code. A duplicate name on DEFINE is condition code 8, as the condition-code table says",
2003 basis: Basis::Chosen,
2004 oracle: Oracle::EnterpriseCobol,
2005 },
2006 Assumption {
2007 id: PRINT_LISTING_LAYOUT,
2008 claim: "IDCAMS PRINT lists as the Access Method Services samples show (z/OS 3.1, idai200/dgt3i239-dgt3i241, da6i2249): LISTING OF DATA SET -name, each record under KEY OF RECORD - (hexadecimal for DUMP and HEX, characters for CHARACTER, da6i2245), RBA OF RECORD - or RECORD SEQUENCE NUMBER -. DUMP lines are 116 columns: a four-digit offset, 32 bytes in groups of four with a wider gap after 16, characters between asterisks; HEX lines are 120 digits, CHARACTER lines 120 characters after a blank line. Characters are the PN chain's (idai200/parm), others print as periods. IDC0005I counts the records listed (ieam600/idc0005i). FROMKEY starts at its key or the next higher, TOKEY stops at its key or the next lower, a key ending X'5C' is generic; a key longer than the data set's ends with IDC3310I (m009223), a key on a data set without keys with IDC3311I (m009224); an empty cluster fails OPEN with 160 (idad500/x1cb): IDC3300I, IDC3351I, condition code 12 (idai200/ccodes)",
2009 basis: Basis::Documented,
2010 oracle: Oracle::EnterpriseCobol,
2011 },
2012 Assumption {
2013 id: PRINT_RANGE_ENDS,
2014 claim: "ironwork chooses: a blank line between records; CHARACTER's blank line and RBA's hyphen from dgt3i239 over da6i2249; RELATIVE RECORD NUMBER - n, empty slots unlisted; an RBA as the sum of earlier record lengths, without control interval boundaries; PN as PL/I's 60-character set, recalled, so lower case prints as periods; a short key generic without the asterisk too; FROMKEY with COUNT and SKIP with TOKEY allowed; SKIP and COUNT through a path in alternate key order. FROMKEY above every key or SKIP past the end ends with IDC3006I, code 12 (m009121); key errors follow IDC3302I; errors replace the listing; a listing of no records (an empty sequential data set, TOKEY below the start, COUNT(0)) ends with IDC0005I 0 and code 4",
2015 basis: Basis::Chosen,
2016 oracle: Oracle::EnterpriseCobol,
2017 },
2018 Assumption {
2019 id: DBCS_UNDER_SINGLE_BYTE_PAGE,
2020 claim: "The Programming Guide has a program with DBCS data items or DBCS literals compiled with one of the mixed CCSIDs of its Table 47 (SC27-8714-03, p. 354), and does not say what a single-byte CODEPAGE does with them. Under a single-byte page ironwork compiles DBCS items and moves and compares their bytes as under a mixed one, since those operations convert nothing; where DBCS data becomes characters, in DISPLAY, JSON GENERATE, a move to a national item, a comparison with one, and NATIONAL-OF, the DBCS space X'4040' is U+3000 and every other character U+FFFD. A DBCS literal, whose bytes only the DBCS component can give, ends the run with abend code IRONWORK where it is used, in ironwork's words, as an alphanumeric literal the page cannot encode does",
2021 basis: Basis::Chosen,
2022 oracle: Oracle::EnterpriseCobol,
2023 },
2024 Assumption {
2025 id: DBCS_LITERAL_SOURCE,
2026 claim: "A DBCS literal's opening delimiter is followed by a shift-out and its closing one preceded by a shift-in (Language Reference SC27-8713-03, p. 41), which a source held in Unicode, as ironwork reads it, no longer carries: ironwork takes the characters between G' (or G\", or N' and N\" under NSYMBOL(DBCS)) and the closing delimiter as the literal, removing a shift-out (U+000E) after the opening delimiter and a shift-in (U+000F) before the closing one where they are present, and encodes each character with the DBCS component of the CODEPAGE, two code points that one DBCS code stands for taking that code. A literal of no character, or of more than 28 (p. 42), is refused when read",
2027 basis: Basis::Chosen,
2028 oracle: Oracle::EnterpriseCobol,
2029 },
2030 Assumption {
2031 id: MIXED_PAGE_DATA,
2032 claim: "Under a mixed CODEPAGE an alphanumeric literal may hold DBCS characters, delimited by shift-out and shift-in (Language Reference SC27-8713-03, pp. 39-40), and alphanumeric data converted to characters (DISPLAY, NATIONAL-OF, a move to a national item) is read the same way: a single byte is the single-byte component's character, and the bytes between X'0E' and the next X'0F' are DBCS characters, two bytes each, the shifts themselves no characters. Encoding prefers the single-byte character and opens a shift-out run only for a character the single bytes lack. A single byte the component leaves unassigned is U+001A, as IBM's conversions substitute, an unassigned DBCS code U+FFFD, and an odd byte left before a shift-in U+FFFD; Enterprise COBOL's conversion of such bytes is not documented",
2033 basis: Basis::Chosen,
2034 oracle: Oracle::EnterpriseCobol,
2035 },
2036 Assumption {
2037 id: DBCS_DISPLAY,
2038 claim: "DISPLAY transfers a DBCS item to the output device with shift-out and shift-in around it (Language Reference SC27-8713-03, p. 333); ironwork writes DISPLAY's line as Unicode text, where a DBCS item or literal is its characters through the CODEPAGE's DBCS component and the shift codes, which are not characters, are not written",
2039 basis: Basis::Chosen,
2040 oracle: Oracle::EnterpriseCobol,
2041 },
2042 Assumption {
2043 id: DBCS_HOST_VARIABLES,
2044 claim: "A DBCS item of PICTURE G or N without B is a GRAPHIC host variable of its characters, and a group of a 49-level binary halfword and a 49-level such item VARGRAPHIC, its length counting DBCS characters, as Db2 declares them; an item whose PICTURE has B has no SQL type. A GRAPHIC value reaches PostgreSQL, or a recording, as the text the CODEPAGE's DBCS component gives its characters, and text comes back as DBCS characters padded with DBCS spaces or cut at a character, SQLWARN1 and the indicator taking the length in characters; under a single-byte CODEPAGE, or for a character the DBCS component lacks, the value is SQLCODE -330, as one the code page cannot convert. Db2's own GRAPHIC conversion between CCSIDs is not modelled",
2045 basis: Basis::Chosen,
2046 oracle: Oracle::Db2,
2047 },
2048 Assumption {
2049 id: INSPECT_SIGNED_ZONED,
2050 claim: "INSPECT examines a signed zoned item as if it had been moved to an unsigned zoned item of its length: an overpunched sign is read as its digit, and a separate sign is not examined and not replaced (Language Reference SC27-8713-03, p. 359, Table 40). The table says REPLACING and CONVERTING copy their result back, and not what becomes of an overpunched sign. ironwork keeps it: a byte at the sign's place that a phrase changes to a digit takes the old sign half, and one no phrase changes is left as it was. The unsigned image changes only the sign byte's zone, as an alphanumeric image of the item does",
2051 basis: Basis::Chosen,
2052 oracle: Oracle::EnterpriseCobol,
2053 },
2054 Assumption {
2055 id: RELATIVE_NUMBER_BELOW_ONE,
2056 claim: "A relative record number below 1 names no record area, the first being number 1 (Language Reference SC27-8713-03, p. 147). A random WRITE with one reports 24, as a write beyond the file's boundaries, and a random READ, REWRITE or DELETE 23, as a record that does not exist. Table 34 (pp. 300-301) gives the meanings of 23 and 24 and does not name a record number below 1",
2057 basis: Basis::Chosen,
2058 oracle: Oracle::EnterpriseCobol,
2059 },
2060 Assumption {
2061 id: DISPLAY_NUMERIC_FUNCTION,
2062 claim: "DISPLAY of an integer or numeric intrinsic function is refused when compiled (S): such a function can be used only where an arithmetic expression can be specified (Language Reference SC27-8713-03, p. 499; Programming Guide SC27-8714-03, p. 56), and DISPLAY's operands are identifiers and literals (p. 333). The manuals give neither the message number nor its text: the message is ironwork's, and the severity is the one C190 gives INSPECT of such a function. MAX and MIN, whose type follows their arguments, are refused when the first is numeric (pp. 591, 599), and CONTENT-OF is not. Nor is a user-defined function: the Language Reference lets a numeric one be used wherever an arithmetic expression can be (p. 77) without saying only there, and DISPLAY shows its value as its RETURNING item",
2063 basis: Basis::Documented,
2064 oracle: Oracle::EnterpriseCobol,
2065 },
2066 Assumption {
2067 id: BY_VALUE_TO_REFERENCE,
2068 claim: "A COBOL program CALLed with an argument BY VALUE whose formal parameter is received BY REFERENCE: the Language Reference requires BY VALUE for both the argument and the parameter (SC27-8713-03, p. 322) and does not say what happens when they differ. On z/OS the parameter list then holds the value where the called program reads an address, and the result turns on what that value addresses. ironwork binds the parameter to storage of its own holding the value, as it does for a parameter received BY VALUE, and the called program runs",
2069 basis: Basis::Chosen,
2070 oracle: Oracle::EnterpriseCobol,
2071 },
2072 Assumption {
2073 id: FLOAT_EXPONENTIATION,
2074 claim: "An exponent with decimal places, or one holding a division or an exponentiation when dmax is above zero, makes its expression floating point, as a floating-point operand or function does (Programming Guide SC27-8714-03, pp. 796, 800). ironwork takes an operand's decimal places from its description, a function's only from a user-defined function's RETURNING item, and dmax as the statement's or the evaluated expression's. A floating-point exponentiation is evaluated in long precision, extended under ARITH(EXTEND) (p. 800); the manuals do not give Language Environment's algorithm, and ironwork gives the value nearest the exact power, computed as the floating-point functions are (C110): an integer exponent by repeated squaring, any other as e^(y ln |x|). Zero to a positive power is zero. Table 32 of the Language Reference (SC27-8713-03, pp. 296-297) gives the rest: zero to a negative power is a size error, and without ON SIZE ERROR the program ends abnormally, which ironwork does with the HFP divide exception, S0CF, a division by zero raises; zero to the power zero is 1, and a negative number to a fractional power is computed with the base's absolute value, each with a message, when no SIZE ERROR phrase is written. ironwork gives those two values whether or not ON SIZE ERROR is written, issues no message, and does not run the phrase for them",
2075 basis: Basis::Chosen,
2076 oracle: Oracle::EnterpriseCobol,
2077 },
2078];
2079
2080pub fn get(id: &str) -> &'static Assumption {
2081 ASSUMPTIONS.iter().find(|a| a.id == id).unwrap_or_else(|| panic!("no assumption {id}"))
2082}
2083
2084pub fn c_series() -> impl Iterator<Item = (usize, &'static Assumption)> {
2087 ASSUMPTIONS.iter().enumerate().map(|(i, a)| (i + 1, a))
2088}
2089
2090#[cfg(test)]
2091mod tests {
2092 use super::*;
2093
2094 #[test]
2095 fn ids_are_unique() {
2096 let mut ids: Vec<_> = ASSUMPTIONS.iter().map(|a| a.id).collect();
2097 ids.sort();
2098 ids.dedup();
2099 assert_eq!(ids.len(), ASSUMPTIONS.len());
2100 }
2101
2102 #[test]
2103 fn c_series_numbers_run_from_one_without_gaps() {
2104 let numbers: Vec<usize> = c_series().map(|(n, _)| n).collect();
2105 assert_eq!(numbers, (1..=ASSUMPTIONS.len()).collect::<Vec<_>>());
2106 }
2107
2108 #[test]
2109 fn c_series_keeps_every_id_once_and_leaves_the_register_alone() {
2110 let before: Vec<&str> = ASSUMPTIONS.iter().map(|a| a.id).collect();
2111 let seen: Vec<&str> = c_series().map(|(_, a)| a.id).collect();
2112 assert_eq!(seen, before);
2113 assert_eq!(ASSUMPTIONS.iter().map(|a| a.id).collect::<Vec<_>>(), before);
2114 }
2115}