rm2k-lib 1.2.0

Reads and writes the RPG Maker 2000/2003 LCF binary formats (LDB, LMU, LMT, LSD).
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
//! Position tracking and chunk framing over a borrowed byte slice.
//!
//! Every LCF chunk is length-prefixed, so [`ChunkIter`] slices each field's payload as a
//! `&'a [u8]` bounded to that chunk before the field sees it. A field that misreads its payload
//! cannot reach the next chunk's bytes: malformed input is contained, not just detected.
//!
//! Every read goes through `slice::get`, returning `None` at end of input. No `get_unchecked` or
//! `transmute` anywhere in the decode path: the input is untrusted, and a mispredicted branch
//! costs less than a CVE.

use crate::{
    diag::{DiagKind, DiagSink, Diagnostic},
    engine::Engine,
    wire::{self, Varint, Varint64},
};

/// Parse-time context threaded through every read: the target engine, and where to report
/// recoverable problems.
///
/// It also carries the string-list entry budget for the load. Reuse one `Ctx` across several loads and
/// they share what the earlier ones left, so a fresh `Ctx` per load is the usual pattern.
///
/// The sink is `dyn` on purpose: diagnostics are the cold path, so the indirect call is free in
/// practice, while monomorphising ~70 generated `read` methods per sink type would bloat code.
/// marshal-rs makes the same trade for its symbol table.
pub struct Ctx<'d> {
    pub engine: Engine,
    diag: &'d mut dyn DiagSink,
    entries: EntryBudget<'d>,
}

/// Most string-list entries one load may produce across every list it reads: 1,048,576 entries, 16 MiB
/// on a 64-bit host. Real files sit far below it. The bound stops a crafted file from expanding
/// loading without limit, including by repeating a chunk.
pub(crate) const ENTRY_BUDGET: u64 = 1 << 20;

/// The entries a load may still produce. The root `Ctx` owns the count. A reborrowed one borrows its
/// parent's, so every nested struct draws on the same budget.
enum EntryBudget<'d> {
    Root(u64),
    Nested(&'d mut u64),
}

impl EntryBudget<'_> {
    fn left(&mut self) -> &mut u64 {
        match self {
            Self::Root(left) => left,
            Self::Nested(left) => left,
        }
    }
}

impl<'d> Ctx<'d> {
    pub fn new(engine: Engine, diag: &'d mut dyn DiagSink) -> Self {
        Self {
            engine,
            diag,
            entries: EntryBudget::Root(ENTRY_BUDGET),
        }
    }

    /// Take `count` entries from the load's string-list budget. Returns `false`, taking nothing, when
    /// fewer remain. Only the `alloc`-gated string lists spend the budget.
    #[cfg(feature = "alloc")]
    #[inline]
    pub(crate) fn take_entries(&mut self, count: u64) -> bool {
        let left = self.entries.left();
        if count > *left {
            return false;
        }
        *left -= count;
        true
    }

    /// Reborrow for a nested struct.
    ///
    /// Schema depth is fixed by the format (four levels at most: `Database` -> `Troop` ->
    /// `TroopPage` -> `EventCommand`), so recursive descent is safe. It would not be for an
    /// attacker-controlled shape.
    #[inline]
    pub fn reborrow(&mut self) -> Ctx<'_> {
        Ctx {
            engine: self.engine,
            diag: &mut *self.diag,
            entries: EntryBudget::Nested(self.entries.left()),
        }
    }

    /// Record a recoverable problem.
    #[inline]
    pub fn report(&mut self, struct_name: &'static str, chunk_id: u32, offset: u32, kind: DiagKind) {
        self.diag.push(Diagnostic {
            offset,
            chunk_id,
            struct_name,
            kind,
        });
    }

    /// Whether the sink has asked parsing to stop (strict mode).
    #[must_use]
    #[inline]
    pub fn should_abort(&self) -> bool {
        self.diag.should_abort()
    }

    /// The latched diagnostic, if the sink latches one.
    #[must_use]
    #[inline]
    pub fn latched(&self) -> Option<Diagnostic> {
        self.diag.first()
    }
}

impl core::fmt::Debug for Ctx<'_> {
    fn fmt(&self, f: &mut core::fmt::Formatter<'_>) -> core::fmt::Result {
        f.debug_struct("Ctx")
            .field("engine", &self.engine)
            .finish_non_exhaustive()
    }
}

/// A position within a borrowed buffer.
///
/// `base` is the offset of `buf[0]` within the *whole* file, so a cursor over a sliced-out chunk
/// payload still reports file-absolute offsets in diagnostics.
#[derive(Clone, Debug)]
pub struct Cursor<'a> {
    buf: &'a [u8],
    pos: usize,
    base: u32,
}

impl<'a> Cursor<'a> {
    /// A cursor over a whole file.
    #[must_use]
    #[inline]
    pub const fn new(buf: &'a [u8]) -> Self {
        Self { buf, pos: 0, base: 0 }
    }

    /// A cursor over a slice that begins `base` bytes into the file.
    #[must_use]
    #[inline]
    pub const fn with_base(buf: &'a [u8], base: u32) -> Self {
        Self { buf, pos: 0, base }
    }

    /// Bytes consumed so far.
    #[must_use]
    #[inline]
    pub const fn pos(&self) -> usize {
        self.pos
    }

    /// File-absolute offset of the current position.
    #[must_use]
    #[inline]
    pub const fn offset(&self) -> u32 {
        self.base.wrapping_add(self.pos as u32)
    }

    /// The whole buffer, consumed part included.
    ///
    /// Needed to slice out a span the cursor has already walked past - the event-command codec
    /// scans its parameters to find their extent, then hands back that exact subslice.
    #[must_use]
    #[inline]
    pub const fn buf(&self) -> &'a [u8] {
        self.buf
    }

    /// The not-yet-consumed bytes.
    #[must_use]
    #[inline]
    pub fn remaining(&self) -> &'a [u8] {
        // `pos` never exceeds `buf.len()`: every advance goes through a checked `take`.
        self.buf.get(self.pos..).unwrap_or(&[])
    }

    /// How many bytes are left.
    #[must_use]
    #[inline]
    pub const fn remaining_len(&self) -> usize {
        self.buf.len() - self.pos
    }

    /// Whether the cursor is exhausted.
    #[must_use]
    #[inline]
    pub const fn is_empty(&self) -> bool {
        self.pos >= self.buf.len()
    }

    /// Consume and return the next `n` bytes, or `None` if fewer remain.
    #[must_use]
    #[inline]
    pub fn take(&mut self, n: usize) -> Option<&'a [u8]> {
        let end = self.pos.checked_add(n)?;
        let out = self.buf.get(self.pos..end)?;
        self.pos = end;
        Some(out)
    }

    /// Consume and return everything left.
    #[must_use]
    #[inline]
    pub fn take_rest(&mut self) -> &'a [u8] {
        let out = self.remaining();
        self.pos = self.buf.len();
        out
    }

    /// Advance by at most `n` bytes, saturating at the end.
    #[inline]
    pub fn skip(&mut self, n: usize) {
        self.pos = self.pos.saturating_add(n).min(self.buf.len());
    }

    /// Consume a compressed integer.
    ///
    /// Advances past the whole encoding even when it overflows the 5-byte cap, matching
    /// `LcfReader::ReadInt`. Returns `None` at end of input.
    #[must_use]
    #[inline]
    pub fn varint(&mut self) -> Option<Varint> {
        let v = wire::decode_varint(self.remaining())?;
        self.pos += v.consumed;
        Some(v)
    }

    /// Consume a compressed integer, substituting `0` on malformed or truncated input.
    ///
    /// This is the lenient form the generated readers use: liblcf yields `0` in both cases, and
    /// real files rely on it, so the value is reported alongside a diagnostic rather than an error.
    #[inline]
    pub fn varint_lenient(&mut self, ctx: &mut Ctx<'_>, struct_name: &'static str, chunk_id: u32) -> i32 {
        let offset = self.offset();
        match self.varint() {
            Some(v) => {
                if v.overflow {
                    ctx.report(struct_name, chunk_id, offset, DiagKind::BadVarint);
                }
                v.value as i32
            }
            None => {
                self.pos = self.buf.len();
                ctx.report(struct_name, chunk_id, offset, DiagKind::ShortChunk);
                0
            }
        }
    }

    /// Consume a 64-bit compressed integer.
    #[must_use]
    #[inline]
    pub fn varint64(&mut self) -> Option<Varint64> {
        let v = wire::decode_varint64(self.remaining())?;
        self.pos += v.consumed;
        Some(v)
    }

    /// Consume one raw byte.
    #[must_use]
    #[inline]
    pub fn u8(&mut self) -> Option<u8> {
        self.take(1).map(|b| b[0])
    }

    /// Consume a little-endian `i16`.
    #[must_use]
    #[inline]
    pub fn i16_le(&mut self) -> Option<i16> {
        self.take(2).map(|b| i16::from_le_bytes([b[0], b[1]]))
    }

    /// Consume a little-endian `u32`.
    #[must_use]
    #[inline]
    pub fn u32_le(&mut self) -> Option<u32> {
        self.take(4).map(|b| u32::from_le_bytes([b[0], b[1], b[2], b[3]]))
    }

    /// Consume a little-endian `i32`.
    #[must_use]
    #[inline]
    pub fn i32_le(&mut self) -> Option<i32> {
        self.take(4).map(|b| i32::from_le_bytes([b[0], b[1], b[2], b[3]]))
    }

    /// Consume a little-endian `f64`.
    #[must_use]
    #[inline]
    pub fn f64_le(&mut self) -> Option<f64> {
        self.take(8)
            .map(|b| f64::from_le_bytes([b[0], b[1], b[2], b[3], b[4], b[5], b[6], b[7]]))
    }

    /// Peek at the next byte without consuming it.
    #[must_use]
    #[inline]
    pub fn peek(&self) -> Option<u8> {
        self.buf.get(self.pos).copied()
    }

    /// Read a length-prefixed byte string: a compressed integer length followed by that many
    /// raw bytes.
    ///
    /// The bytes are returned **verbatim**. This crate never transcodes: LCF carries no encoding
    /// marker of any kind, so guessing would be exactly that. See the crate docs.
    #[must_use]
    #[inline]
    pub fn lstring(&mut self) -> Option<&'a [u8]> {
        let len = self.varint()?;
        self.take(len.value as usize)
    }
}

/// One framed chunk: `varint(id) || varint(len) || len bytes`.
#[derive(Clone, Copy, Debug)]
pub struct Chunk<'a> {
    pub id: u32,
    pub data: &'a [u8],
    pub offset: u32,
}

impl<'a> Chunk<'a> {
    /// A cursor over this chunk's payload, reporting file-absolute offsets.
    #[must_use]
    #[inline]
    pub const fn cursor(&self) -> Cursor<'a> {
        Cursor::with_base(self.data, self.offset)
    }
}

/// Walks a chunk stream until the terminating zero id or the end of the buffer.
///
/// This is not an [`Iterator`]: advancing needs a [`Ctx`] to report a malformed frame.
#[derive(Clone, Debug)]
pub struct ChunkIter<'a> {
    cur: Cursor<'a>,
    done: bool,
}

impl<'a> ChunkIter<'a> {
    /// Frame the chunks in `buf`, whose first byte sits at file offset `base`.
    #[must_use]
    #[inline]
    pub const fn new(buf: &'a [u8], base: u32) -> Self {
        Self {
            cur: Cursor::with_base(buf, base),
            done: false,
        }
    }

    /// Frame the chunks in a cursor's remaining bytes.
    #[must_use]
    #[inline]
    pub const fn from_cursor(cur: Cursor<'a>) -> Self {
        Self { cur, done: false }
    }

    /// The underlying cursor, positioned after the last chunk yielded.
    #[must_use]
    #[inline]
    pub const fn cursor(&self) -> &Cursor<'a> {
        &self.cur
    }

    /// Recover the cursor, positioned just past the terminating zero id.
    ///
    /// Needed by the indexed-list codec: a list's elements are chunk streams laid end to end
    /// inside one payload with no length prefix of their own, so after reading an element the
    /// caller must resume from wherever that element's terminator left off.
    #[must_use]
    #[inline]
    pub const fn into_cursor(self) -> Cursor<'a> {
        self.cur
    }

    /// Bytes left unread after the stream terminated.
    #[must_use]
    #[inline]
    pub fn tail(&self) -> &'a [u8] {
        self.cur.remaining()
    }

    /// Consume a whole chunk stream, returning the span it occupied.
    ///
    /// A borrowed iterator would have to be driven to the end before moving on, and silently
    /// desynchronises if the caller stops early. Returning the span lets a caller re-walk the
    /// element any number of times, for the cost of a skim over the chunk headers.
    #[must_use]
    pub fn consume(cur: &mut Cursor<'a>, ctx: &mut Ctx<'_>, struct_name: &'static str) -> &'a [u8] {
        let start = cur.pos();
        let mut chunks = Self::from_cursor(cur.clone());
        while chunks.next(ctx, struct_name).is_some() {}
        *cur = chunks.into_cursor();
        cur.buf().get(start..cur.pos()).unwrap_or(&[])
    }

    /// Find one chunk in a stream by id, without allocating.
    ///
    /// `O(chunks)`, which is the only option: a chunk stream has no index and its entries are
    /// variable-width. Fine for reaching one known field; walk the stream yourself if you want
    /// several.
    #[must_use]
    pub fn find(body: &'a [u8], base: u32, id: u32, ctx: &mut Ctx<'_>, struct_name: &'static str) -> Option<Chunk<'a>> {
        let mut chunks = Self::new(body, base);
        while let Some(chunk) = chunks.next(ctx, struct_name) {
            if chunk.id == id {
                return Some(chunk);
            }
        }
        None
    }

    /// Yield the next chunk.
    ///
    /// Returns `None` on the terminating zero id, at end of input, or once a frame has been found
    /// unrecoverable. A chunk whose declared length runs past the end of the buffer is truncated
    /// to what is actually there and reported as [`DiagKind::ChunkOverrun`].
    #[must_use]
    pub fn next(&mut self, ctx: &mut Ctx<'_>, struct_name: &'static str) -> Option<Chunk<'a>> {
        if self.done || self.cur.is_empty() {
            return None;
        }

        let id_offset = self.cur.offset();
        let id = match self.cur.varint() {
            Some(v) if v.overflow => {
                ctx.report(struct_name, 0, id_offset, DiagKind::BadVarint);
                self.done = true;
                return None;
            }
            Some(v) => v.value,
            None => {
                self.done = true;
                return None;
            }
        };

        // Id 0 terminates the struct (`src/reader_struct_impl.h:68`).
        if id == 0 {
            self.done = true;
            return None;
        }

        let len_offset = self.cur.offset();
        let len = match self.cur.varint() {
            Some(v) if v.overflow => {
                ctx.report(struct_name, id, len_offset, DiagKind::BadVarint);
                self.done = true;
                return None;
            }
            Some(v) => v.value as usize,
            None => {
                ctx.report(struct_name, id, len_offset, DiagKind::ShortChunk);
                self.done = true;
                return None;
            }
        };

        let offset = self.cur.offset();
        let data = match self.cur.take(len) {
            Some(data) => data,
            None => {
                ctx.report(struct_name, id, offset, DiagKind::ChunkOverrun);
                self.done = true;
                self.cur.take_rest()
            }
        };

        Some(Chunk { id, data, offset })
    }
}

/// One element of an indexed list.
#[derive(Clone, Copy, Debug)]
pub struct ListItem<'a> {
    /// The element's id, from the compressed integer preceding its chunk stream.
    pub id: i32,
    /// The element's chunk stream, terminating zero included.
    pub body: &'a [u8],
    /// File-absolute offset of `body`.
    pub offset: u32,
}

impl<'a> ListItem<'a> {
    /// The element's chunks.
    #[must_use]
    #[inline]
    pub const fn chunks(&self) -> ChunkIter<'a> {
        ChunkIter::new(self.body, self.offset)
    }

    /// One of the element's chunks, by id.
    #[must_use]
    #[inline]
    pub fn find(&self, id: u32, ctx: &mut Ctx<'_>, struct_name: &'static str) -> Option<Chunk<'a>> {
        ChunkIter::find(self.body, self.offset, id, ctx, struct_name)
    }
}

/// Walks an indexed list without allocating.
///
/// The core-tier counterpart to [`List<T>`](crate::field::List), which needs `alloc` because it
/// materialises every element. This yields borrowed spans instead, so an embedded consumer can
/// reach one map's events, or one actor's name, without a heap.
#[derive(Clone, Debug)]
pub struct ListWalker<'a> {
    cur: Cursor<'a>,
    left: u32,
    name: &'static str,
}

impl<'a> ListWalker<'a> {
    /// Start walking the list in a chunk payload.
    ///
    /// `struct_name` names the *element* type, and is only used to locate diagnostics.
    #[must_use]
    pub fn new(body: &'a [u8], base: u32, ctx: &mut Ctx<'_>, struct_name: &'static str) -> Self {
        let mut cur = Cursor::with_base(body, base);
        let count = cur.varint_lenient(ctx, struct_name, 0);
        // Each element costs at least two bytes (an id and a terminator), which bounds a count
        // that came from a hostile file.
        let left = (count.max(0) as usize).min(cur.remaining_len() / 2 + 1) as u32;
        Self {
            cur,
            left,
            name: struct_name,
        }
    }

    /// How many elements the list declared and has not yielded yet.
    #[must_use]
    #[inline]
    pub const fn remaining(&self) -> u32 {
        self.left
    }

    /// Yield the next element.
    #[must_use]
    pub fn next(&mut self, ctx: &mut Ctx<'_>) -> Option<ListItem<'a>> {
        if self.left == 0 || self.cur.is_empty() {
            return None;
        }
        self.left -= 1;

        let id = self.cur.varint_lenient(ctx, self.name, 0);
        let offset = self.cur.offset();
        let body = ChunkIter::consume(&mut self.cur, ctx, self.name);
        Some(ListItem { id, body, offset })
    }
}