Bug 381185 – sys-apps/coreutils-8.12: tests fail inside of the sandbox

• Status: RESOLVED NEEDINFO
• Product: Gentoo Linux

[Description Toralf Förster 2011-08-30 15:57:32 UTC]

I complained about a failed test case (pls see URL) and got this answer (therefore I filed this bug report) :

> FAIL: dd/skip-seek-past-dev (exit: 1)
> =====================================
> ...
> + env SANDBOX_WRITE=:/dev/console:/dev/fd:/dev/full:/dev/null:/dev/pts/:/dev/pty:/dev/shm:/dev/stderr:/dev/stdout:/dev/tts:/dev/tty:/dev/vc/:/dev/zero:/proc/self/fd:/tmp/:/usr/lib32/cf:/usr/lib32/conftest:/usr/lib64/cf:/usr/lib64/conftest:/usr/lib/cf:/usr/lib/conftest:/usr/tmp/cf:/usr/tmp/conftest:/var/tmp:/var/tmp/:/var/tmp/ccache:/var/tmp/portage/sys-apps/coreutils-8.12/homedir/.bash_history:/dev/full:/dev/root timeout 10 dd bs=1 seek=41047646721 count=0 status=noxfer
> ./dd/skip-seek-past-dev: line 51: /dev/root: Permission denied

The above looks suspiciously like someone has modified this test script.
In particular, the "env SANDBOX_WRITE=..." bit is not a part of the
distributed script.  It seems to be something gentoo specific.



Reproducible: Always

[Comment 1 Toralf Förster 2011-09-13 11:00:45 UTC]

Created attachment 286329 [details]
/var/log/portage/sys-apps:coreutils-8.13:20110913-101613.log.gz

In version 8.13 there are few more failed test cases with the Gentoo's patched version, whereas the original unmodified version passes all tests.

[Comment 2 SpanKY 2011-09-15 04:13:41 UTC]

yes, but are you running `make check` inside of sandbox ?  or are you manually running it ?

the changes made to the tests for sandbox only *enable* access for a few device nodes.  the /dev/root perm denied should not be caused by that.

further, always post `emerge --info` with your bug reports.

[Comment 3 Toralf Förster 2011-09-15 08:26:02 UTC]

(In reply to comment #2)
> yes, but are you running `make check` inside of sandbox ?  or are you manually
> running it ?
I did manually run it both as user root and as non-root user within the unpacked straight tar-ball directory both with "./configure" and with the same configure line as portage.

[Comment 4 SpanKY 2011-09-17 23:08:45 UTC]

(In reply to comment #3)

so it's only failing inside of the sandbox.  the patch we apply merely exports a variable so it's highly unlikely it has anything to do with the failure.

[Comment 5 Toralf Förster 2011-09-18 09:38:51 UTC]

(In reply to comment #4)
> so it's only failing inside of the sandbox.
Yes (tested 8.13 + Gentoo patches outside of portage - all tests passed).