[Pkg-samba-maint] Bug#558453: Bug#558453: Bug#558453: 558453: samba memory leak (fix not til 3.4.5)

[Ross Boylan]

On Wed, 2010-01-06 at 18:16 +0100, Christian PERRIER wrote:
> > Certainly. However, we (pkg maintainers) are not really in position to
> > help that much here, except by acting as proxies with upstream.
> > 
> > Are you in position to report this in upstream's Bugzilla
> > (https://bugzilla.samba.org) so that you can investigate this deeper
> > directly with them?
> 
> 
> I have to say: WOW.. Upstream bug #7020, which you opened, is really
> impressive and I wanted to thank you publicly for the great
> cooperation you have with upstream in investigating it. I'll be more
> than happy to close #558453 when we release 3.4.4 packages.
> 
You're welcome; it's been a lot of work.

The latest comments on the bug indicate the fix will be part of 3.4.5,
not 3.4.4.  It might be worth applying the patches to earlier versions,
given the seriousness of the problem.  I think the problem is much more
acute with Vista printing than with other clients, because it generates
a lot of useless messages.  It sounds as if fixes for that problem might
even be in 3.4.4 (see https://bugzilla.samba.org/show_bug.cgi?id=6754)

Also, it's possible some of the other problems reported in this bug have
other causes.  On the other hand, Debian bug 538819 may be the same
thing.  It's not clear if the valgrind output, even with
--trace-children=yes, in that bug was sufficient to catch the real leak.
I only caught it after I killed the main samba process.

Note also that Message 40 of 538819 refers to a problem in samba 3.2;
some of the comments in upstream 7020 indicate the problems there might
be introduced more recently (e.g., comment 44 says "this is also an area
that changed between 3.3.x and 3.4.x").

Ross