Re: Xorg 7.2

[David Nusinow <dnusinow@speakeasy.net>]

On Thu, Apr 19, 2007 at 04:13:35AM -0700, Steve Langasek wrote:
> On Wed, Apr 18, 2007 at 09:15:21PM -0400, David Nusinow wrote:
> > Also of major note, Julien clarified my post a bit in the comments. Most of
> > 7.2 is actually in unstable, including the drivers (in addition, we have
> > pre-release drivers for intel in experimental). The missing thing though is
> > the server (which is, of course, the centerpiece of the release), which
> > will go in with the coming version 1.3 release, with or without XCB. We
> > need to push XCB forward though, and how to deal with the java bug
> > mentioned in that post isn't clear yet.
> 
> I'm totally confused by this.  I thought a way forward had been identified
> in our discussion on IRC, has something changed?

It got more complicated after that. I uploaded the XCB NMU with the Novell
patch to provide an environment variable that the user could set to allow
the locking check to fail in XCB without terminating the program. The idea
was then to patch our java package to allow java to run, but also for us to
find and fix the bugs elsewhere. We really do want to catch those bugs now
and I don't want to wait for Sun to get off their asses so that we can do
it.

Looking at the java package, it'd take a bit of complicated reworking to
wrap the java binary in a script that sets the environment variable.
Feasible, but a little invasive for a band-aid rather than a real fix, so I
haven't done it yet.

On top of that, Jaime and Josh (XCB guys, for those following along at
home) were surprised that the java apps would work at all with XCB even
with the lock check disabled. I haven't extensively tested java apps
running with XCB yet to see if it'll work consistently.

What I really would like to do is actually fix the bug in java though. We
spent some time hunting down the java source and trying to figure out the
problem. Apparently the java binaries that are distributed by Sun
statically link old versions of the X libs that have locking bugs (now
fixed) in them. Preloading the libs doesn't seem to work on the more recent
versions of java though, so there may well be a bug in the code. I'm hoping
to have some time to figure out where the bug is and fix it, but that's
going to require figuring out a few basic things first.

If anyone has any mysterious contacts at Sun that they could talk to about
this, that'd be great. They've got the bug report in their database, and
since the whole "java-in-debian" thing was a totally backroom affair, I
have no idea who to contact about this to ask them to put some real effort
in to it.

 - David Nusinow