decodecorpus #1664
decodecorpus #1664
Conversation
Add test case for short bistream
|
@terrelln |
|
A targeted test case for an extreme outlier, |
tests/zstreamtest.c
Outdated
| CHECK_Z( ZSTD_decompressStream(zds, &outBuff, &inBuff) ); | ||
| } | ||
|
|
||
| { XXH64_state_t xxhStateIn, xxhStateOut; |
There was a problem hiding this comment.
nit:
if all you want to do is compare decompressed and decodedBuffer content and ensure they are equal, you could simply memcmp() them.
tests/zstreamtest.c
Outdated
| outBuff.pos = 0; | ||
|
|
||
| while (inBuff.pos < inBuff.size) { | ||
| CHECK_Z( ZSTD_decompressStream(zds, &outBuff, &inBuff) ); |
There was a problem hiding this comment.
nit: it's slightly stronger than CHECK_Z() :
presuming the frame is complete (is it ?), you want a return value 0, to indicate that the decoder has reached the end of frame. You also want to ensure it has consumed the entire input (inBuff.pos == inBuff.compressedSize).
There was a problem hiding this comment.
Yes, it's a complete frame.
|
Looks good ! |
reflect code review comments
Decoder was enforcing a constraint that is more strict than the spec. Sequence section for a compressed block can be less than the constraint that was enforced by decoder. When a block with rle mode for all (literal length, match length, and offset) is followed by another block with repeat mode for all (literal length, match length, and offset), the second block's sequence section can be 2 bytes (1 byte for number of sequences and 1 byte for compression mode for each field).
This pull request includes 2 changes.
Run
./decodecorpus -ptestfiles -otestfiles -n1 -s1236947365 -vto generate an input that has an issue before the fix.