nptl/test-*-printers tests failing

nptl/test-*-printers tests failing

[Steve Ellcey]

Does anyone else get failures for the new nptl/test-*-printers tests?
On aarch64, with the standard sources I get these failures:

FAIL: nptl/test-cond-printers
FAIL: nptl/test-condattr-printers
FAIL: nptl/test-mutex-printers
FAIL: nptl/test-mutexattr-printers
FAIL: nptl/test-rwlock-printers
FAIL: nptl/test-rwlockattr-printers

When I ran the testsuite on x86_64 I got:

UNSUPPORTED: nptl/test-cond-printers
UNSUPPORTED: nptl/test-condattr-printers
UNSUPPORTED: nptl/test-mutex-printers
UNSUPPORTED: nptl/test-mutexattr-printers
UNSUPPORTED: nptl/test-rwlock-printers
UNSUPPORTED: nptl/test-rwlockattr-printers

I looked at the *.out files on my aarch64 system but there is no output
there and the test-result files just say 'original exit status 127'.

Before I dig into this some more I was wondering if anyone else was
seeing these failures or knew what was causing them?

Steve Ellcey
sellcey@caviumnetworks.com

Re: nptl/test-*-printers tests failing

[Steve Ellcey]

On Fri, 2016-12-16 at 14:48 -0800, Steve Ellcey wrote:
> Does anyone else get failures for the new nptl/test-*-printers tests?
> On aarch64, with the standard sources I get these failures:
> 
> FAIL: nptl/test-cond-printers
> FAIL: nptl/test-condattr-printers
> FAIL: nptl/test-mutex-printers
> FAIL: nptl/test-mutexattr-printers
> FAIL: nptl/test-rwlock-printers
> FAIL: nptl/test-rwlockattr-printers

And just after sending the email I found the problem.  On the machine
I was testing there is no /usr/bin/python.  There is /usr/bin/python3
which is why I could run build-many-glibcs.py but there was no
/usr/bin/python.

I found this in my log file:

env PYTHONPATH=/home/ubuntu/sellcey/glibc-patched/obj-
glibc64/nptl/:../scripts:${PYTHONPATH} \
    python test-mutexattr-printers.py test-mutexattr-printers.c
/home/ubuntu/sellcey/glibc-patched/obj-glibc64/nptl/test-mutexattr-
printers nptl-printers.py > /home/ubuntu/sellcey/glibc-patched/obj-
glibc64/nptl/test-mutexattr-printers.out; \
../scripts/evaluate-test.sh nptl/test-mutexattr-printers $? false false
> /home/ubuntu/sellcey/glibc-patched/obj-glibc64/nptl/test-mutexattr-
printers.test-result
env: ‘python’: No such file or directory


There should probably be some sort of check for the existence of the
python command as part of these tests.


Steve Ellcey
sellcey@caviumnetworks.com

Re: nptl/test-*-printers tests failing

[Siddhesh Poyarekar]

On Saturday 17 December 2016 05:24 AM, Steve Ellcey wrote:
> There should probably be some sort of check for the existence of the
> python command as part of these tests.

I'll add a configure test for this tonight.

Siddhesh

Re: nptl/test-*-printers tests failing

[Martin Galvan]

2016-12-16 20:54 GMT-03:00 Steve Ellcey <sellcey@caviumnetworks.com>:
> And just after sending the email I found the problem.  On the machine
> I was testing there is no /usr/bin/python.  There is /usr/bin/python3
> which is why I could run build-many-glibcs.py but there was no
> /usr/bin/python.

Does that apply to the UNSUPPORTED tests as well? There are a few
things that can make the tests fail with that code, namely an
old/absent gdb/python/pexpect, etc. Do the .out files show you
anything on x86_64?

Re: nptl/test-*-printers tests failing

[Steve Ellcey]

On Sat, 2016-12-17 at 12:37 -0300, Martin Galvan wrote:
> 2016-12-16 20:54 GMT-03:00 Steve Ellcey <sellcey@caviumnetworks.com>:
> > 
> > And just after sending the email I found the problem.  On the machine
> > I was testing there is no /usr/bin/python.  There is /usr/bin/python3
> > which is why I could run build-many-glibcs.py but there was no
> > /usr/bin/python.
> Does that apply to the UNSUPPORTED tests as well? There are a few
> things that can make the tests fail with that code, namely an
> old/absent gdb/python/pexpect, etc. Do the .out files show you
> anything on x86_64?

Yes, the x86 out files contain:

% more test-condattr-printers.out
PExpect 4.0 or newer must be installed to test the pretty printers.


I don't have pexpect installed, but at least with this problem I get
UNSUPPORTED and not FAIL.

Steve Ellcey
sellcey@caviumnetworks.com

Re: nptl/test-*-printers tests failing

[Martin Galvan]

Alright. I saw that Sid wrote a new patch for picking the python
binary at configure time, could you give it a try?

https://sourceware.org/ml/libc-alpha/2016-12/msg00677.html

Thanks!

Re: nptl/test-*-printers tests failing

[Steve Ellcey]

On Mon, 2016-12-19 at 13:15 -0300, Martin Galvan wrote:
> Alright. I saw that Sid wrote a new patch for picking the python
> binary at configure time, could you give it a try?
> 
> https://sourceware.org/ml/libc-alpha/2016-12/msg00677.html
> 
> Thanks!

The patch did not work for me.  I did get UNSUPPORTED when python
existed (because of no pexpect), but when I moved python and python3
out of the way so that the configure/build/test could not find them:

checking for python3... no
checking for python... no
configure: WARNING:
*** These auxiliary programs are missing or incompatible versions:
msgfmt python
*** some features or tests will be disabled.
*** Check the INSTALL file for required versions.


The printer tests still showed up as FAIL:

FAIL: nptl/test-cond-printers
FAIL: nptl/test-condattr-printers
FAIL: nptl/test-mutex-printers
FAIL: nptl/test-mutexattr-printers
FAIL: nptl/test-rwlock-printers
FAIL: nptl/test-rwlockattr-printers

I don't see anything in the patch where these tests are not run when
PYTHON is not found.  Is something missing from the patch as posted?
Maybe a change to nptl/Makefile?

Steve Ellcey
sellcey@caviumnetworks.com

Re: nptl/test-*-printers tests failing

[Siddhesh Poyarekar]

On Monday 19 December 2016 11:58 PM, Steve Ellcey wrote:
> I don't see anything in the patch where these tests are not run when
> PYTHON is not found.  Is something missing from the patch as posted?
> Maybe a change to nptl/Makefile?

That is odd, you shouldn't need a change in nptl/Makefile because the
target is disabled in Rules when PYTHON does not exist.  I have to admit
though that I did not test the case where python did not exist because I
did not have that kind of a minimal setup - maybe the PYTHON variable is
being set incorrectly.  Can you tell me what it is set to in your
$(common-objpfx)/config.make in all those cases?

I'll set up a chroot anyway and fix up.

Thanks,
Siddhesh

Re: nptl/test-*-printers tests failing

[Steve Ellcey]

On Tue, 2016-12-20 at 10:15 +0530, Siddhesh Poyarekar wrote:
> On Monday 19 December 2016 11:58 PM, Steve Ellcey wrote:
> > 
> > I don't see anything in the patch where these tests are not run
> > when
> > PYTHON is not found.  Is something missing from the patch as
> > posted?
> > Maybe a change to nptl/Makefile?
> That is odd, you shouldn't need a change in nptl/Makefile because the
> target is disabled in Rules when PYTHON does not exist.  I have to
> admit
> though that I did not test the case where python did not exist
> because I
> did not have that kind of a minimal setup - maybe the PYTHON variable
> is
> being set incorrectly.  Can you tell me what it is set to in your
> $(common-objpfx)/config.make in all those cases?
> 
> I'll set up a chroot anyway and fix up.
> 
> Thanks,
> Siddhesh

# Script execution tools.
BASH = /bin/bash
AWK = gawk
PERL = /usr/bin/perl
PYTHON = no

I did my testing by just moving /usr/bin/python to /usr/bin/python.old
and /usr/bin/python3 to /usr/bin/python3.old and then doing the build
and test.

Steve Ellcey
sellcey@caviumnetworks.com

Re: nptl/test-*-printers tests failing

[Siddhesh Poyarekar]

On Tuesday 20 December 2016 09:34 PM, Steve Ellcey wrote:
> PYTHON = no

Hah, that's what I had guessed.  I'll fix that *and* test it this time.
Martin also pointed out another little goof-up off-list, which I'll add.

Thanks,
Siddhesh

Re: nptl/test-*-printers tests failing

[Martin Galvan]

Here's the new patch, could you try this Steve?

https://sourceware.org/ml/libc-alpha/2016-12/msg00822.html

Thanks.

Re: nptl/test-*-printers tests failing

[Steve Ellcey]

On Wed, 2016-12-21 at 18:09 -0300, Martin Galvan wrote:
> Here's the new patch, could you try this Steve?
> 
> https://sourceware.org/ml/libc-alpha/2016-12/msg00822.html
> 
> Thanks.

It works.  It seems odd to have the tests show up as UNSUPPORTED
when python exists but the needed modules don't exist but have the
tests not run at all (and thus not show up at all) when python itself
does not exist.  I understand from an implementation point of view why
it is done that way and I guess I wouldn't reject the patch for that
reason, but it does strike me as odd.  But the configure does warn
about features/tests being disabled so maybe that is OK.

configure: WARNING:
*** These auxiliary programs are missing or incompatible versions: msgfmt python
*** some features or tests will be disabled.
*** Check the INSTALL file for required versions.

Steve Ellcey
sellcey@caviumnetworks.com

Re: nptl/test-*-printers tests failing

[Siddhesh Poyarekar]

On Thursday 22 December 2016 04:08 AM, Steve Ellcey wrote:
> On Wed, 2016-12-21 at 18:09 -0300, Martin Galvan wrote:
>> Here's the new patch, could you try this Steve?
>>
>> https://sourceware.org/ml/libc-alpha/2016-12/msg00822.html
>>
>> Thanks.
> 
> It works.  It seems odd to have the tests show up as UNSUPPORTED
> when python exists but the needed modules don't exist but have the
> tests not run at all (and thus not show up at all) when python itself
> does not exist.  I understand from an implementation point of view why
> it is done that way and I guess I wouldn't reject the patch for that
> reason, but it does strike me as odd.  But the configure does warn
> about features/tests being disabled so maybe that is OK.

I've posted v3 that marks the tests as unsupported instead of ignoring them.

Siddhesh