[debian-mysql] Bug#660686: ooops

[Nicholas Bamber]

severity 660686 grave
thanks

Severity raised at request of release team.


(14:18:35) jcristau: there hasn't been a libpq SONAME bump in a while
(14:18:39) jcristau: so i don't know
(14:18:47) periapt: mm
(14:21:16) jcristau: hmm, libpq doesn't have versioned symbols.  bad libpq.
(14:26:25) ***periapt mumbles something about #660686
(14:28:43) jcristau: periapt: hrm.  that's *bad*.
(14:29:26) jcristau: i think this should be grave, not normal.
(14:29:45) periapt: Ooops
(14:30:30) periapt: jcristau: Implications?
(14:31:20) periapt: If we patched it that would be  non-upstream SONAME 
bump, right?
(14:32:06) jcristau: hmm?
(14:46:15) periapt: jcristau: If it is so serious why is  there not a 
lintian check for it?
(14:46:38) jcristau: lintian can't check for regressions
(14:48:00) periapt: Ah it's serious because it's a regression.
(14:48:52) jcristau: and because libmysqlclient has half the world as 
reverse deps
(14:52:17) periapt: jcristau: So if we could patch it would you want us to?
(14:53:11) jcristau: yes
(14:53:34) periapt: And that would mean another transition?
(14:55:40) jcristau: periapt: no, it would only bump shlibs
(14:56:16) jcristau: bah, stupid build system
(14:59:13) periapt: jcristau: Showing my ignorance here: 
http://bugs.mysql.com/bug.php?id=64386 the symbols in objdump are 
"different".
(15:00:03) jcristau: periapt: a reference to my_connect at Base can be 
satisfied by my_connect at libmysqlclient_18
(15:00:20) periapt: jcristau: Thanks
(15:01:32) periapt: jcristau: I'll reply to the bug report but I would 
rather upload 5.5.24 to wheezy before looking at it. I can experiment in 
our main line and backport.