[arch-dev-public] Integrity Check x86_64

[Xavier]

On Sun, Oct 5, 2008 at 2:23 PM, Roman Kyrylych <roman.kyrylych at gmail.com> wrote:
> 2008/10/3  <repomaint at archlinux.org>:
>> Warning: Ensure your ABS tree is clean to prevent false positives.
>>
>> ==========================
>> = x86_64 Integrity Check =
>> ==========================
>>
>> Performing integrity checks...
>> ==> parsing pkgbuilds
>> ==> checking dependencies
>> ==> checking makedepends
>> ==> checking archs
>> ==> checking for circular dependencies
>> ==> checking repo packages
>>
>> Missing Dependencies
>> ----------------------
>> libgdiplus --> 'libungif'
>> klibc-module-init-tools --> '/lib/klibc-*.so'
>> klibc-module-init-tools --> '.so'
>> latex2html --> 'libungif'
>> imlib --> 'libungif'
>> v86d --> '/lib/klibc-*.so'
>> v86d --> '.so'
>> klibc-extras --> '/lib/klibc-*.so'
>> klibc-extras --> '.so'
>> klibc-udev --> '/lib/klibc-*.so'
>> klibc-udev --> '.so'
>> windowmaker --> 'libungif'
>> klibc-kbd --> '/lib/klibc-*.so'
>> klibc-kbd --> '.so'
>
> I think there's a bug in the script ;-)
>

There is a bug indeed, the problem is that the current python script
used a rather quite complicated method, it was trying to parse the
PKGBUILD bash scripts itself, in a naive way, handling variables and
stuff, but not everything and it broke in the klibc case.

But even if it worked, it would still be wrong : it is simply
impossible to check the klibc dependencies at the PKGBUILD level,
because the dependency is computed at runtime. For the klibc package
itself, it looks at the library just built (pkg/lib/klibc-*.so). For
the other klibc-* packages, it looks at the library installed on the
system (/lib/klibc-*.so).
In short, you can check the klibc* dependencies at the package level,
or at the sync databases level, examining the /var/lib/pacman/sync/
contents, or the *.db.tar.gz databases. But not at the PKGBUILD level.

Anyway, because of the complex parsing and other reasons, I totally
rewrote the check-archlinux script, and Aaron committed it one month
ago : http://projects.archlinux.org/?p=dbscripts.git;a=commit;h=909b017c08109bda405a3e38a59cbf26211e6683
In particular, this new script includes a hack to skip the klibc-*
dependency. However this hack was not fully working, so I posted a
patch to improve it recently :
http://www.archlinux.org/pipermail/arch-dev-public/2008-September/008310.html
This patch was not pulled yet.

Otherwise, Aaron told me that the problem in pushing all these
dbscript changes live is that it will require him one day of
availability, so that he can fix if anything goes wrong.