[ANNOUNCE] xorg-server 1.10.2

Cyril Brulebois kibi at debian.org
Tue May 31 01:36:26 PDT 2011


Jeremy Huddleston <jeremyhu at freedesktop.org> (30/05/2011):
> My RC2 release notes indicated that I didn't expect to include any
> changes before the final release.  If I had known that these were
> regressions, I surely would've pushed to get these changes into
> master.  In order to avoid communication problems like this in the
> future, I've updated the wiki with instructions on how to
> communicate "Do Not Release" issues:
> http://www.x.org/wiki/Server110Branch


> I was under the impression that the "crash after setting root
> background pixmap to None, then setting color" issue was not a
> regression.

It's not, it's just a “nice to have” at some point. I just wanted to
point out that with that mail I tried to draw the other bug to your

> Regarding #36986, I was unfortunately unaware of the scope of the
> issue.  I was under the impression that this was a broken test
> rather than a broken server.  I've since CC'd myself on that bug and
> marked it as a blocker for xserver-1.10.

As confirmed by Peter, it's “just a broken test”. But the net result
is: the testsuite no longer passes.

(What follows is quite general, not specific to the xserver.)

Having spent some years within Debian (with an eye on Ubuntu and other
distros), reactions to a broken testsuite can be:
 1) let's disable it entirely (yuck),
 2) let's run it but ignore the outcome (meaning somebody will have to
    check all build logs, which usually doesn't happen, so yuck),
 3) let's disable the broken tests (OK-ish).

I can't speak for other packagers, but having a suddenly failing test
suite can mean a bit of digging to achieve 3), and I'm afraid 1) or 2)
could be picked instead (e.g. because there's some kind of hurry to
get a package to build again, for some transition or release
purposes). All in all, I find it a bit sad to potentially lower QA.

(FWIW, I picked:

Workaround fdo#36986 <http://bugs.freedesktop.org/show_bug.cgi?id=36986>
--- a/test/input.c
+++ b/test/input.c
@@ -1213,8 +1213,10 @@
         g_assert(v->numAxes == num_axes);
+#if 0
         g_assert(((void*)v->axisVal - (void*)v) % sizeof(double) == 0);
         g_assert(((void*)v->axes - (void*)v) % sizeof(double) == 0);
         num_axes ++;

> If these issues are as big as you are making them seem, then I'm
> inclined to shorten the 1.10.3 release cycle.

I guess I wouldn't care so much if it had been a test failure on say
itanium. The fact it happens on x86(-32, not -64) makes me find it a
bit more troublesome. But if you're OK with a failing test suite, I
guess there's no reason to shorten the 1.10.3 release cycle.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.x.org/archives/xorg-devel/attachments/20110531/029402ba/attachment.pgp>

More information about the xorg-devel mailing list