pbzip2 -cd sometimes (rarely) truncates data

[Kevin]

I occasionally get Gentoo binary packages that will not unpack. I am aware of the previous issues with trailing garbage but this appears to be different and--ignore-trailing-garbage does not affect this problem...

If I do 'tar -tvf broken.tar.bz2' it works fine.
If I do 'pbzip2 -d broken.tar.bz2' I get a correct tar file.
However, if I do 'pbzip2 -cd broken.tar.bz2 > broken.tar' I get a tar file that is truncated at 16384 bytes.

I have placed a small (40KB) example file here for testing: https://sanitarium.net/broken.tar.bz2

This was tested on version 1.1.12 on a Gentoo Linux system.

When I hit a package file that has this problem it is repeatable with that file. My only solutions when I hit this problem are to either delete and rebuilt the package or to temporarily remove the bzip2>pbzip2 symlink so that portage will use regular bzip2 which is fine with these files.

[Yavor Nikolov (yavor-nikolov)]

Thanks for reporting that issue. I managed to reproduce it on Ubuntu.

Another way I managed to get a non-truncated tar: pbzip2 -cd < broken.tar.bz2 > broken.tar

[Kevin (m-launch3ad-m)]

Thanks for testing and confirming.

I wasn't really looking for alternative command lines as I would normally just do a tar -xvf but my problem is that portage is doing something like:
pbzip2 -d -cq broken.tar.bz2 | tar -xp -C /tmp/.... -f -
Which also fails. The >broken.tar syntax was the obvious most simple case that both demonstrated the problem and showed that the data was being truncated at an interesting number of bytes.

[Yavor Nikolov (yavor-nikolov)]

Hi Kevin,
I just pushed a fix to lp:~pbzip2/pbzip2/bug-1524909-cd-truncates-output-sometimes. Would be great if you can test it. (If it's OK we can proceed with merging and releasing the fix).

[Kevin (m-launch3ad-m)]

Hello Yavor,

This did fix the problem with the file and more importantly it allowed Gentoo's portage system to merge the binary package correctly.

Thanks for the help.