[lldb-dev] Code coverage of llvm, clang, lldb & polly

[Malea, Daniel]

Cool stuff. I found and fixed another that your packaging builds detected;
the crashes in #173
<http://llvm-jenkins.debian.net/job/llvm-toolchain-quantal-binaries/archite
cture=amd64,distribution=quantal/173/> should be resolved in the next
build.

That said, there's still a few test cases that don't build correctly on
the Jenkins bots. Is there any way we can grab a directory of logs from
the build? Specifically, seeing the contents of the
llvm-build/lldb-test-traces directory would be tremendous help.





Thanks,Dan



On 2013-06-01 11:35 AM, "Sylvestre Ledru" <sylvestre at debian.org> wrote:

>http://buildd-clang.debian.net/coverage/
>Your patch fixed the issue!
>Thanks to it, the LLVM toolchain is close to the 80 % lines test
>coverage (which is excellent).
>
>Sylvestre
>
>
>
>On 31/05/2013 18:04, Malea, Daniel wrote:
>> Yeah, it's somewhat surprising that 'ps' is not being found..it doesn't
>> seem to be on the PATH. In any case, no rush, I will continue looking at
>> some of the other test failures that are happening only on Debian builds
>> in the mean time..
>> 
>> Thanks,
>> Dan
>> 
>> On 2013-05-31 11:58 AM, "Sylvestre Ledru" <sylvestre at debian.org> wrote:
>> 
>>> Oh. I would have say that ps was installed by default in a chroot :)
>>> Maybe it was(is?) the case for Ubuntu.
>>> Anyway, I just applied your patch and relaunched a build (but note that
>>> the code coverage takes
>>> between 4h to 5h30)
>>>
>>> Thanks :)
>>> Sylvestre
>>>
>>> Le 31/05/2013 17:51, Malea, Daniel a écrit :
>>>> Thanks for the update. It's clear that ps is not being installed in
>>>>the
>>>> chroot environments. I checked the debian/control file and there's no
>>>> build depends on it; try the attached patch that adds a Build-Depends
>>>>on
>>>> procps package.
>>>>
>>>> Thanks,
>>>> Dan
>>>>
>>>>
>>>> On 2013-05-31 10:46 AM, "Sylvestre Ledru" <sylvestre at debian.org>
>>>>wrote:
>>>>
>>>>> On 31/05/2013 00:31, Malea, Daniel wrote:
>>>>>> Hi Sylvestre, I committed a potential fix for the 'ps' issue in
>>>>>> 182965.
>>>>> It fails with:
>>>>> Traceback (most recent call last):
>>>>>  File
>>>>>
>>>>> 
>>>>>"/tmp/buildd/llvm-toolchain-snapshot-3.4~svn182989/lldb/test/dotest.py
>>>>>",
>>>>> line 1210, in <module>
>>>>>    print >> f, "Command invoked: %s\n" % getMyCommandLine()
>>>>>  File
>>>>>
>>>>> 
>>>>>"/tmp/buildd/llvm-toolchain-snapshot-3.4~svn182989/lldb/test/dotest.py
>>>>>",
>>>>> line 1074, in getMyCommandLine
>>>>>    ps = subprocess.Popen([which('ps'), '-o', "command=CMD",
>>>>> str(os.getpid())], stdout=subprocess.PIPE).communicate()[0]
>>>>>  File "/usr/lib/python2.7/subprocess.py", line 711, in __init__
>>>>>    errread, errwrite)
>>>>>  File "/usr/lib/python2.7/subprocess.py", line 1308, in
>>>>>_execute_child
>>>>>    raise child_exception
>>>>> AttributeError: 'NoneType' object has no attribute 'rfind'
>>>>>
>>>>>
>>>>>> Out of curiosity, is the code-coverage run happening on a different
>>>>>> machine than the normal testing runs? For example, I'm seeing:
>>>>>>
>>>>>>
>>>>>>
>>>>>> 
>>>>>>http://llvm-jenkins.debian.net/job/llvm-toolchain-quantal-binaries/ar
>>>>>>ch
>>>>>> it
>>>>>> ec
>>>>>> ture=amd64,distribution=quantal/166/consoleText
>>>>>>
>>>>>> Doesn't have any problems finding 'ps' when running the tests. It
>>>>>> seems
>>>>>> odd that 'ps' is not being found..could it maybe be in a (different)
>>>>>> chroot environment where ps is not available for some reason?
>>>>> Besides the code coverage build flags, there is no difference.
>>>>> However, all the system is different. That means that the Python
>>>>>under
>>>>> quantal is probably different from the one in Debian Unstable.
>>>>>
>>>>> Cheers,
>>>>> Sylvestre
>>>>>
>>>
>> 
>
>_______________________________________________
>lldb-dev mailing list
>lldb-dev at cs.uiuc.edu
>http://lists.cs.uiuc.edu/mailman/listinfo/lldb-dev