girt 0.17.0

An incremental, idiomatic Rust implementation of Git
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

//! EWAH framing is checked without expanding runs; decoding has a separate shared work budget.
use std::sync::atomic::AtomicBool;

use super::{Error, bad, charge, check, tail_mask};

#[derive(Clone, Debug)]
pub(super) struct Ewah {
    start: usize,
    words: usize,
}

impl Ewah {
    pub(super) fn parse(
        bytes: &[u8],
        cursor: &mut usize,
        end: usize,
        objects: usize,
    ) -> Result<Self, Error> {
        let bits = word(bytes, *cursor)? as usize;
        let words = word(bytes, cursor.checked_add(4).ok_or(bad("EWAH offset"))?)? as usize;
        if bits.div_ceil(64) > objects.div_ceil(64) {
            return Err(bad("EWAH bit length"));
        }
        let start = cursor.checked_add(8).ok_or(bad("EWAH offset"))?;
        let data_end = words
            .checked_mul(8)
            .and_then(|length| start.checked_add(length))
            .ok_or(bad("EWAH words"))?;
        let next = data_end.checked_add(4).ok_or(bad("EWAH trailer"))?;
        if words == 0 || next > end {
            return Err(bad("EWAH range"));
        }
        let current = word(bytes, data_end)? as usize;
        let mut position = 0;
        let mut expanded = 0_usize;
        let mut last_rlw = 0;
        let mut last_word = 0;
        while position < words {
            last_rlw = position;
            let header = long(bytes, start + position * 8)?;
            position += 1;
            let run = ((header >> 1) & 0xffff_ffff) as usize;
            let literals = (header >> 33) as usize;
            expanded = expanded
                .checked_add(run)
                .and_then(|count| count.checked_add(literals))
                .ok_or(bad("EWAH expansion"))?;
            if expanded > bits.div_ceil(64) {
                return Err(bad("EWAH expansion"));
            }
            if run != 0 {
                last_word = if header & 1 != 0 { u64::MAX } else { 0 };
            }
            position = position
                .checked_add(literals)
                .filter(|&next| next <= words)
                .ok_or(bad("EWAH literal range"))?;
            if literals != 0 {
                last_word = long(bytes, start + (position - 1) * 8)?;
            }
        }
        if expanded != bits.div_ceil(64) || current != last_rlw {
            return Err(bad("EWAH length or current run"));
        }
        if expanded != 0 && last_word & !tail_mask(bits) != 0 {
            return Err(bad("EWAH trailing bits"));
        }
        if expanded != 0 && expanded == objects.div_ceil(64) && last_word & !tail_mask(objects) != 0
        {
            return Err(bad("bitmap outside object universe"));
        }
        *cursor = next;
        Ok(Self { start, words })
    }

    pub(super) fn xor_into(
        &self,
        bytes: &[u8],
        output: &mut [u64],
        budget: &mut usize,
        cancel: &AtomicBool,
    ) -> Result<(), Error> {
        let mut position = 0;
        let mut out = 0;
        while position < self.words {
            check(cancel)?;
            let header = long(bytes, self.start + position * 8)?;
            position += 1;
            let run = ((header >> 1) & 0xffff_ffff) as usize;
            let literals = (header >> 33) as usize;
            charge(budget, run + literals)?;
            let run_end = out + run;
            if header & 1 != 0 {
                for word in output
                    .get_mut(out..run_end)
                    .ok_or(bad("decoded run range"))?
                {
                    check(cancel)?;
                    *word ^= u64::MAX;
                }
            }
            out = run_end;
            for _ in 0..literals {
                check(cancel)?;
                let value = long(bytes, self.start + position * 8)?;
                *output.get_mut(out).ok_or(bad("decoded literal range"))? ^= value;
                out += 1;
                position += 1;
            }
        }
        Ok(())
    }
}

pub(super) fn short(bytes: &[u8], start: usize) -> Result<u16, Error> {
    let end = start.checked_add(2).ok_or(bad("integer offset"))?;
    let value = bytes.get(start..end).ok_or(bad("integer range"))?;
    Ok(u16::from_be_bytes(
        value.try_into().map_err(|_| bad("integer width"))?,
    ))
}

pub(super) fn word(bytes: &[u8], start: usize) -> Result<u32, Error> {
    let end = start.checked_add(4).ok_or(bad("integer offset"))?;
    let value = bytes.get(start..end).ok_or(bad("integer range"))?;
    Ok(u32::from_be_bytes(
        value.try_into().map_err(|_| bad("integer width"))?,
    ))
}

pub(super) fn long(bytes: &[u8], start: usize) -> Result<u64, Error> {
    let end = start.checked_add(8).ok_or(bad("integer offset"))?;
    let value = bytes.get(start..end).ok_or(bad("integer range"))?;
    Ok(u64::from_be_bytes(
        value.try_into().map_err(|_| bad("integer width"))?,
    ))
}

#[cfg(test)]
mod tests {
    use super::*;

    fn encoded(bits: u32, words: &[u64], current: u32) -> Vec<u8> {
        let mut bytes = Vec::new();
        bytes.extend_from_slice(&bits.to_be_bytes());
        bytes.extend_from_slice(&(words.len() as u32).to_be_bytes());
        for word in words {
            bytes.extend_from_slice(&word.to_be_bytes());
        }
        bytes.extend_from_slice(&current.to_be_bytes());
        bytes
    }

    #[test]
    fn runs_literals_and_xor_decode_with_shared_budget() {
        // Two one-words followed by a literal, then one zero-word.
        let bytes = encoded(256, &[1 | (2 << 1) | (1 << 33), 5, 1 << 1], 2);
        let ewah = Ewah::parse(&bytes, &mut 0, bytes.len(), 256).unwrap();
        let mut output = [0; 4];
        ewah.xor_into(&bytes, &mut output, &mut 4, &AtomicBool::new(false))
            .unwrap();
        assert_eq!(output, [u64::MAX, u64::MAX, 5, 0]);
        ewah.xor_into(&bytes, &mut output, &mut 4, &AtomicBool::new(false))
            .unwrap();
        assert_eq!(output, [0; 4]);
        assert!(matches!(
            ewah.xor_into(&bytes, &mut output, &mut 3, &AtomicBool::new(false)),
            Err(Error::Limit(_))
        ));
    }

    #[test]
    fn rejects_invalid_framing_before_expansion() {
        for (bits, words, current, objects) in [
            (64, vec![u64::MAX], 0, 64),  // Huge run and literal lengths.
            (64, vec![1 << 33], 0, 64),   // Missing literal.
            (64, vec![1 << 1], 1, 64),    // Current RLW points outside the buffer.
            (1, vec![1 << 33, 2], 0, 1),  // Bit outside declared size.
            (64, vec![1 << 33, 2], 0, 1), // Bit outside actual inventory.
            (128, vec![1 << 1], 0, 128),  // Short expansion.
            (0, vec![], 0, 0),            // No RLW.
        ] {
            let bytes = encoded(bits, &words, current);
            assert!(Ewah::parse(&bytes, &mut 0, bytes.len(), objects).is_err());
        }
    }
}