Cyril Brulebois
2024-11-04 23:00:01 UTC
Reply
PermalinkVersion: 2:1.8.10-1
Severity: serious
Tags: d-i
Justification: broken shlibs, breaks udebs, etc.
X-Debbugs-Cc: debian-***@lists.debian.org
Hi,
[ Now turning the initial mail ping into a proper bug report. ]
Spotted via dose's reporting lots of uninstallable udebs on arm64: the
new shlibs for libx11-6 is broken, with various libx* packages now
depending on libx11-xcb1 (not a udeb).
Loading Image...
The new debian/rules has:
override_dh_makeshlibs:
dh_makeshlibs -a -plibx11-6 -V'libx11-6 (>= 2:1.6.0)' --add-udeb=libx11-6-udeb -- -c4
dh_makeshlibs -a -plibx11-xcb1 -V'libx11-xcb1' -- -c4
dh_makeshlibs -a -Nlibx11-6 -Nlibx11-xcb1 -- -c4
which leads to the following for libx11-6.shlibs (amd64):
libX11 6 libx11-xcb1
udeb: libX11 6 libx11-xcb1
Meanwhile, libx11-xcb1 has:
libX11-xcb 1 libx11-xcb1
which seems to match the intent in changelog.
Out of curiosity, pausing the build after the first dh_makeshlibs call,
the shlibs for libx11-6 is indeed correct at that point:
libX11 6 libx11-6 (>= 2:1.6.0)
udeb: libX11 6 libx11-6-udeb (>= 2:1.6.0)
and the second call is responsible for busting it up, since afterwards
it's amended to become what's found in the package:
libX11 6 libx11-xcb1
udeb: libX11 6 libx11-xcb1
So it looks like some confusion due to the -a/-p/-N combinations?
Going back to pausing between first and second dh_makeshlibs calls,
here's the slibs file for libx11-xcb1…
libX11-xcb 1 libx11-6 (>= 2:1.6.0)
udeb: libX11-xcb 1 libx11-6 (>= 2:1.6.0)
which convinces me the flag combinations aren't fine as they are.
Cheers,
--
Cyril Brulebois (***@debian.org) <https://debamax.com/>
D-I release manager -- Release team member -- Freelance
Cyril Brulebois (***@debian.org) <https://debamax.com/>
D-I release manager -- Release team member -- Freelance