bug in gipfeli

[inikep]

In this excerpt one can see that input_block is overridden so prev_block becames pointer to a free memory. It causes crashes on files longer than 64KB especially on Windows.

gipfeli-internal.cc:

string input_block;
...
prev_block = input_block.data();
input_block = block_reader.GetNextBlock();

BTW, Reporting issues for gipfeli on GitHub is turned off.

[nemequ]

Reporting issues may be off, but you can still file pull requests. This sounds like https://github.com/google/gipfeli/pull/4

Regardless, I don't think gipfeli's author hangs out here… If you find an issue but don't want to actually fix it so you can file a PR, I would suggest reporting it via e-mail.

[Jyrki Alakuijala]

Regardless, I don't think gipfeli's author hangs out here…

We are here. I'll ask Rasto to take a look.

[inikep]

Reporting issues may be off, but you can still file pull requests. This sounds like https://github.com/google/gipfeli/pull/4

This is the same problem.

The patch from jibsen works for me, but now I have decompression errors. I think that it may be a requirement of continuity of memory between prev_block and input_block or going outside a buffer during compression.

[nemequ]

We are here. I'll ask Rasto to take a look.

Nice, thanks. Are you involved in Gipfeli at all, or do you just know (work with?) him? I know you're both at Google, but it is a somewhat large company…

The patch from jibsen works for me, but now I have decompression errors. I think that it may be a requirement of continuity of memory between prev_block and input_block or going outside a buffer during compression.

FWIW Jørgen has another PR open against Gipfeli which *might* resolve that. I can't really tell from the description if those are the same issue, but I know he filed both of those PRs in an effort to get it working with Squash on Windows, so the odds are decent.

[inikep]

The second patch solves decompression errors: https://github.com/google/gipfeli/pull/5