fuzz/c/std: enlarge the brotli fuzzer's encoder dst; clarify brcat's parity claim

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UZ978Fjdsyig4v3wYAZRg5
This commit is contained in:
Claude committed 2026-10-27 17:27:38 +00:00
1 parent 7d040276e5
commit c9bb73e3bf
2 files changed
+11 -5

No files matched your search

+7 -4
View File
@@ -22,10 +22,13 @@ Flags:
-quality=N the encoder's quality, 0 to 11. The default is 11.
-fail-if-unsandboxed exit with an error if the seccomp sandbox failed.
The encoder's output is byte-for-byte identical to the brotli reference
implementation's (libbrotlienc v1.2.1) with the same quality, lgwin and large
window parameters, on every platform. Like that reference implementation,
qualities 10 and 11 are much slower than lower qualities.
The encoder's output is byte-for-byte identical, on every platform, to what
the brotli reference implementation's library (libbrotlienc v1.2.1) produces
for the same parameters and the same sequence of calls. That output depends
on how the input is split into calls (for example, the first call's length is
the default size hint), so it can differ from the "brotli" command line tool's,
which reads 512 KiB at a time and sets a size hint. Like that reference
implementation, qualities 10 and 11 are much slower than lower qualities.
To run:
+4 -1
View File
@@ -132,7 +132,10 @@ static const uint8_t g_serialized_dictionary[] = {
// at qualities 10 and 11, which are slow.
#define ENC_MAX_INPUT 65536
#define ENC_MAX_INPUT_HQ 4096
#define ENC_DST_ARRAY_SIZE (2 * ENC_MAX_INPUT + 65536)
// With 1-byte chunks and frequent flushes, each input byte can become its own
// metablock plus a padding block (about 4 bytes per input byte at qualities 0
// and 1), so the dst buffer is sized well above that.
#define ENC_DST_ARRAY_SIZE (8 * ENC_MAX_INPUT + 65536)
static wuffs_brotli__encoder g_encoder;
static wuffs_brotli__decoder g_rt_decoder;