Bug#934132: unblock: elogind/241.3-1+debian1
Package: release.debian.org
Severity: normal
User: release.debian.org@packages.debian.org
Usertags: unblock
Dear Release Team,
Further to recent discussions on #debian-release and at the suggestion of jmw I
request unblocking of the testing transition of elogind.
Transition of elogind to testing was blocked on 16 July after a discussion on
irc.
As far as I understand, the objections raised on irc have been:-
1) To Conflicts/Provides/Replaces libsystemd0 is too dangerous.
2) What prevents this getting installed where it shouldn't.
3) Claims to provide libsystemd ABI, but does not.
4) libelogind0 breaks the example program in sd_journal_next.
So far there have been no reports of breakages relating to 1). We are currently
early in the release cycle. It seems to me that this is precisely the time to be
testing this more widely.
2) is prevented in most situations by libelogind0 also conflicting with
systemd. You cannot coinstall systemd and libelogind0. It is possible to replace
libsystemd0 with libelogind0 in chroots and initless containers. However, my
testing of such environments demonstrates no breakage. In such a scenario the
functions in both libsystemd0 and libelogind0 return non fatal errors.
3) is open to interpretation depending on your definition of ABI
compatibility. Certainly libelogind.so exports the same symbols as libsystemd.so
with identical binary interface. This was the agreed solution to #923244 so that
policykit-1 works with libelogind without recompilation. However, there are
functional differences betweent he two libraries and libelogind0 only provides a
subset of libsystemd0 functionality. The libelogind0 package description
describes these limitations fully although I am happy to reword or make that
more explicit if you feel it could be improved.
4) In the light of the libelogind0 package description saying
"The sd-login(3), sd-bus(3) and sd-id128(3) APIs are implemented in full,
sd-journal(3) is only implemented by redirecting sd_journal_print* and
sd_journal_send* calls to syslog(3) with other functions returning -ENOSYS and
sd-hwdb(3) contains stubs returning -ENOSYS."
the example program in sd_journal_next behaves as you would expect by exiting with
'Function not implemented' in response to sd_journal_open() returning -ENOSYS.
I agree wholeheartedly that Debian's elogind should not break systemd. To date
there have been no reports of any such breakage and I will be happy to receive
and resolve bug reports in any such eventuality.
Elogind provides choice and makes it possible to install many current desktop
environments on non-systemd init systems: I have successfully tested xfce,
cinnamon, mate and budgie and there are reports of gnome working too.
Thank you for your consideration.
Mark
Reply to: