bug#10096: Failing gc.test for i586

bug#10096: Failing gc.test for i586

[Petr Gajdos]

[-- Attachment #1: Type: text/plain, Size: 209 bytes --]

Hi,

gc.test is failing with guile 2.0.3 for me on i586:
FAIL: gc.test: gc: Unused modules are removed

I don't observe this for guile 2.0.2. Could you please 
guide me to resolve this issue?

Petr


[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

bug#10096: Failing gc.test for i586

[Petr Gajdos]

[-- Attachment #1: Type: text/plain, Size: 295 bytes --]

On Mon, Nov 21, 2011 at 03:46:53PM +0100, Petr Gajdos wrote:
> gc.test is failing with guile 2.0.3 for me on i586:
> FAIL: gc.test: gc: Unused modules are removed
> 
> I don't observe this for guile 2.0.2. Could you please 

And it also doesn't happen on x86_64 with guile 2.0.3.

Petr

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

bug#10096: Failing gc.test for i586

[Andy Wingo]

On Mon 21 Nov 2011 15:46, Petr Gajdos <pgajdos@suse.cz> writes:

> gc.test is failing with guile 2.0.3 for me on i586:
> FAIL: gc.test: gc: Unused modules are removed
>
> I don't observe this for guile 2.0.2. Could you please 
> guide me to resolve this issue?

Hummmm.  Well, on the one side, it is not really a fatal failure.  This
tests allocates 1000 modules, then does a few gc runs, then checks to
make sure they were all collected.  Since libgc is a conservative
collector, it is possible that it mistakenly thinks one of those modules
is still alive -- a problem that is more acute on 32-bit systems.  It is
also dependent on the compiler you use, the version of libgc, etc.

I think if I were you, I would just ignore this error.  I know it's not
a very good answer, but does that work for you?

Andy
-- 
http://wingolog.org/

bug#10096: Failing gc.test for i586

[Petr Gajdos]

[-- Attachment #1: Type: text/plain, Size: 999 bytes --]

On Wed, Nov 23, 2011 at 01:06:21PM +0100, Andy Wingo wrote:
> On Mon 21 Nov 2011 15:46, Petr Gajdos <pgajdos@suse.cz> writes:
> 
> > gc.test is failing with guile 2.0.3 for me on i586:
> > FAIL: gc.test: gc: Unused modules are removed
> >
> > I don't observe this for guile 2.0.2. Could you please 
> > guide me to resolve this issue?
> 
> Hummmm.  Well, on the one side, it is not really a fatal failure.  This
> tests allocates 1000 modules, then does a few gc runs, then checks to
> make sure they were all collected.  Since libgc is a conservative
> collector, it is possible that it mistakenly thinks one of those modules
> is still alive -- a problem that is more acute on 32-bit systems.  It is
> also dependent on the compiler you use, the version of libgc, etc.
> 
> I think if I were you, I would just ignore this error.  I know it's not
> a very good answer, but does that work for you?

Andy, 

thanks a lot for you answer, I will ignore it.

Best regards,
Petr

[-- Attachment #2: Type: application/pgp-signature, Size: 198 bytes --]

bug#10096: Failing gc.test for i586

[Andreas Schwab]

threads.test has the same problem: the test "mutex with owner not
retained (bug #27450)" can fail because of a spurious stack reference to
the mutex object.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#10096: Failing gc.test for i586

[Andy Wingo]

On Sat 03 Dec 2011 01:26, Andreas Schwab <schwab@linux-m68k.org> writes:

> threads.test has the same problem: the test "mutex with owner not
> retained (bug #27450)" can fail because of a spurious stack reference to
> the mutex object.

Thanks for the note.  We have a patch that didn't make 2.0.5, but which
might fix this issue for you.  Would you mind trying a snapshot?  Either
from git, or this tarball:

  http://hydra.nixos.org/build/1963047/download/4/guile-2.0.5.3-7e9a3.tar.xz

Thanks!

Andy
-- 
http://wingolog.org/

bug#10096: Failing gc.test for i586

[Andreas Schwab]

Andy Wingo <wingo@pobox.com> writes:

> On Sat 03 Dec 2011 01:26, Andreas Schwab <schwab@linux-m68k.org> writes:
>
>> threads.test has the same problem: the test "mutex with owner not
>> retained (bug #27450)" can fail because of a spurious stack reference to
>> the mutex object.
>
> Thanks for the note.  We have a patch that didn't make 2.0.5, but which
> might fix this issue for you.

gc.test is still failing.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#10096: Failing gc.test for i586

[Andy Wingo]

On Wed 01 Feb 2012 16:16, Andreas Schwab <schwab@linux-m68k.org> writes:

> Andy Wingo <wingo@pobox.com> writes:
>
>> On Sat 03 Dec 2011 01:26, Andreas Schwab <schwab@linux-m68k.org> writes:
>>
>>> threads.test has the same problem: the test "mutex with owner not
>>> retained (bug #27450)" can fail because of a spurious stack reference to
>>> the mutex object.
>>
>> Thanks for the note.  We have a patch that didn't make 2.0.5, but which
>> might fix this issue for you.
>
> gc.test is still failing.

But threads.test passes?  Which gc.test fails?

Thanks,

Andy
-- 
http://wingolog.org/

bug#10096: Failing gc.test for i586

[Andreas Schwab]

Andy Wingo <wingo@pobox.com> writes:

> But threads.test passes?

Due to the nature of these tests the absense of a failure doesn't tell
you much.

> Which gc.test fails?

FAIL: gc.test: gc: Unused modules are removed

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#10096: Failing gc.test for i586

[Andy Wingo]

On Wed 01 Feb 2012 17:13, Andreas Schwab <schwab@linux-m68k.org> writes:

> Andy Wingo <wingo@pobox.com> writes:
>
>> But threads.test passes?
>
> Due to the nature of these tests the absense of a failure doesn't tell
> you much.

Indeed.  As much as I would like for there to be certainty...

>> Which gc.test fails?
>
> FAIL: gc.test: gc: Unused modules are removed

OK, we're back to the original bug.  I added a hack to make this one
throw 'unresolved if it fails.  With that, we should be done here.
Thanks for the feedback.

Andy
-- 
http://wingolog.org/