fix root cause of #3416 #3419
fix root cause of #3416 #3419
Conversation
| ZSTD_memcpy(nextHuf, prevHuf, sizeof(*prevHuf)); | ||
| return ZSTD_compressRleLiteralsBlock(dst, dstCapacity, src, srcSize); | ||
| } | ||
| if ((srcSize >= 8) || allBytesIdentical(src, srcSize)) { |
There was a problem hiding this comment.
Please add a comment as to why this check is necessary.
This is a really hard test to define. |
Ok, it'd be great if we had it, but if it is too hard, we have the fuzzers, and we know that they will quickly catch any regressions. |
A minor change in 5434de0 changed a `<=` into a `<`, and as an indirect consequence allowed compression attempt of literals when there are only 6 literals to compress (previous limit was effectively 7 literals). This is not in itself a problem, as the threshold is merely an heuristic, but it emerged a bug that has always been there, and was just never triggered so far due to the previous limit. This bug would make the literal compressor believes that all literals are the same symbol, but for the exact case where nbLiterals==6, plus a pretty wild combination of other limit conditions, this outcome could be false, resulting in data corruption. Replaced the blind heuristic by an actual test for all limit cases, so that even if the threshold is changed again in the future, the detection of RLE mode will remain reliable.
as requested by @terrelln
A minor update in 5434de0 changed a
<=into a<, and as an indirect consequence allowed compression attempt of literals when there are only 6 literals to compress (previous limit was effectively 7 literals).This is not in itself a problem, as the threshold is merely an heuristic, but it emerged a bug that has always been there, and was just never triggered so far due to the previous limit. This bug would make the literal compressor believes that all literals are the same symbol, but for the exact case where
nbLiterals==6, plus a pretty wild combination of several exceptional conditions, this outcome could be false, resulting in data corruption.Replaced the blind heuristic by an actual test for all limit cases. Even if the threshold is changed again in the future, the detection of Repeated Literal mode will remain reliable.
fix #3416