py-xmethods.exp regression on Fedora Rawhide [Re: [PATCH v3] Fix for PR gdb/17017]

[Jan Kratochvil <jan dot kratochvil at redhat dot com>]

On Wed, 18 Jun 2014 13:40:49 +0200, Siva Chandra wrote:
> On Mon, Jun 16, 2014 at 7:12 AM, Pedro Alves <palves@redhat.com> wrote:
> > On 06/12/2014 07:23 PM, Siva Chandra wrote:
> >> testsuite/
> >> 2014-06-12  Siva Chandra Reddy  <sivachandra@google.com>
> >>
> >>         PR gdb/17017
> >>         * gdb.python/py-xmethods.cc: Add global function call counters and
> >>         increment them in their respective functions.  Remove "cout"
> >>         statements.
> >>         * gdb.python/py-xmethods.exp: Make tests check the global function
> >>         call counters instead of depending on inferior IO.
[...]
> Pushed: 5d376983ca914e1cf36f4968cc87957f9033ebcd

5d376983ca914e1cf36f4968cc87957f9033ebcd is the first bad commit
commit 5d376983ca914e1cf36f4968cc87957f9033ebcd
Author: Siva Chandra <sivachandra@chromium.org>
Date:   Thu Jun 5 07:03:56 2014 -0700
    Make xmethods tests not to depend on inferior IO.

on Fedora 20 x86_64 (python-2.7.5-13.fc20.x86_64):
	p g.size_diff<  unsigned long  >()^M
	From Python G<>::size_diff()^M
	$35 = 4^M
	(gdb) PASS: gdb.python/py-xmethods.exp: After: g.size_diff<unsigned long>()

on Fedora Rawhide x86_64 (python-2.7.7-2.fc21.x86_64):
	p g.size_diff<  unsigned long  >()^M
	Couldn't find method dop::G<int>::size_diff<  unsigned long  >^M
	(gdb) FAIL: gdb.python/py-xmethods.exp: After: g.size_diff<unsigned long>()

I do not know what it depends on.  When I replace the binaries py-xmethods the
regression remains the same so it does not depend on GCC version (4.8 vs. 4.9).

Maybe it depends on Python version?

Before this commit 5d376983ca914e1cf36f4968cc87957f9033ebcd
both Fedora 20 x86_64 and Fedora Rawhide x86_64 were PASSing everything.

Some other logs suggest it happens on all x86_64, x86_64-m32 and i686 but
I cannot guarantee it right now.


Jan