[pkg-wine-party] Bug#535154: Bug#535154: wine: package broken after /emul -> /usr/lib32 transition

[Ove Kaaven]

severity 535154 important
stop

(For other Wine bugs related to the ia32 transition, see #534238,
#533315, and #535097.)

Török Edwin skrev:
> I reported a bug for libc6-i386 that it should have a Breaks: wine, because it
> broke wine during the transition.

There might not be much reason for them to, as only Wine versions in
experimental were affected. Nothing happened to those in lenny and sid
(they use /usr/lib, not /usr/lib32). Also, even if they did care about
experimental, there probably aren't any broken Wine package versions to
add a Breaks line about anyway (see below), so they probably couldn't do
this if they wanted to.

> However wine should be updated to move to /usr/lib32, because now dpkg does
> think that wine has its files in /usr/lib32 (but it didn't move the actual
> files):

If I understand your report correctly, the Wine package you've installed
already uses /usr/lib32. If so, the package is *not* broken.

(Perhaps there's a different version now in the archive which might be
broken, I'm not sure. But yours isn't.)

It is a well-known dpkg problem that it will never automatically convert
symlinks to directories on upgrade (but blindly follow them). The
involved packages must therefore use Breaks and Pre-Depends to ensure
that the directories are fully emptied (i.e., all relevant packages
uninstalled) before the libc upgrade. Then libc's maintainer scripts can
delete the old symlink, and then the updated packages can be installed,
and the files be placed in the new location.

Using Breaks might be tricky in this case, though, since the packages
*already* used the right paths, and did not need to be changed. Which
means there's no package version to use Breaks against.

You have a choice of reinstalling the Wine debs (apt-get install
--reinstall on the affected debs might suffice), or manually moving the
files around to match the dpkg -L output.

I'm not sure it's worth it to add maintainer scripts to either libc or
wine that tries to automate any workarounds, given that the problem is
only in experimental, and never happened in sid. Hopefully people that
use experimental know how to fix their systems when it breaks - at least
know how to reinstall a package.

I will leave the bug open so that other people with similar issues know
what's going on, but there's no need to fix the package. Thus,
downgrading to non-RC.