search

tukaani-project / xz

XZ Utils
https://tukaani.org/xz/
Other
542 stars 101 forks source link

[Bug]: reject a valid lzma file #20

Closed boofish closed 1 year ago

boofish commented 1 year ago

Describe the bug

Hi, I recently fuzz the xz for parsing lzma format file and found one interesting case. Specifically, xz rejects one valid lzma format file while another parser, the 7zip accepts it.

Meanwhile, I use the command ./7zz e -so -tlzma:s0 invalid.lzma for 7zip, and use xz -d -c --format=lzma invalid.lzma for xz.

I have contacted 7zip, it seems that xz should not reject it.

The contact link: https://sourceforge.net/p/sevenzip/bugs/2380/

The interesting file is attached.invalid.zip

Version

5.2.7

Operating System

ubuntu 22.04

Relevant log output

No response

JiaT75 commented 1 year ago

Hi! Thank you for the bug report, but I will close this bug report because it is a documented feature of XZ Utils. 7zip and XZ Utils are almost completely compatible with how they treat .xz and .lzma files, but here is an example of where they differ.

My quick maths determined this file has the settings pb = 1, lp = 3, lc = 5, which is unsupported by XZ Utils. XZ Utils will only compress or decompress .lzma and .xz files if lp + lc <=4.

This is is documented in doc/lzma-file-format.txt (~ line 105 as of 2022-07-13) and src/liblzma/api/lzma/lzma12.h (~ line 280 as of version 5.4.1). I was not part of the project when this decision was made, but my understanding is that files with lc + lp > 4 are unlikely to improve compression significantly and will use a lot more memory and computation time when compressing or decompressing.

Since .lzma has been a legacy format since 2009 and .xz does not support these types of settings, we do not plan to change this. Old .lzma files that have been created with these settings can still be decompressed with 7zip and new files should be using the .xz format anyway.

boofish commented 1 year ago

thanks for your detailed explanation!