[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: dh_auto_test fails and I do not understand why



Quoting Niels Thykier (2022-12-15 10:59:10)
> Long story short:
> 
>   * Bug in fakeroot (#1023286 + #1024544)
>   * Me thinking it was a bug in debhelper so I tried to fix it
>     (which did not work and broke on the way in)
>   * Me realizing it was a bug in fakeroot and my change did not
>     even function as a work around, so I undid it (it broke on
>     the way out as well).
> 
> And the winners are: All the people that have (and are) able to use 
> "Rules-Requires-Root: no" as packages with that flag would have been
> completely unaffected by this entire ordeal!
> 
> Your options include:
> 
>   * Migrate to "Rules-Requires-Root: no" if you can
>   * Ensure you have debhelper (>= 13.11.3~) ("just uploaded") or
>     debhelper/13.11.1 (or debhelper << 13.11)
> 
> I will now go back to looking more at my prototype for getting even more 
> packages buildable with "Rules-Requires-Root: no".

the original cause of all of this is the fakeroot bug since glibc 2.34,
specifically since coreutils was rebuilt with glibc 2.34.

This is a call for help. Please have a look at #1023286 and #1024544.

Back in August, glib 2.34 reached unstable. In October, coreutils 9.1 was
uploaded and used the new glibc functionality. Since then, calls to chown and
other utilities use some new glibc functions on armel, armhf and i386, namely
__stat64_time64, __fstatat64_time64 and __lstat64_time64.

Clint quickly fixed the problem and I submitted a test case so that this kind
of issue does not fly under the radar again in the future. The test case seems
to be working because now we get a build failure on mipsel because the test is
not passing.

So something else is missing. It's likely another glibc function that is
mipsel-specific. I tried to run ltrace on a mipsel porterbox to figure out what
library calls are done by coreutils chown on mipsel but that failed with
"unexpected breakpoint" errors, see #1023436.

Maybe somebody can find out what is wrong with fakeroot on mipsel? Since
fakeroot is a central part of Debian package builds, it would be a shame to
have this bug open for much longer...

Thanks!

cheers, josch

Attachment: signature.asc
Description: signature


Reply to: