[Mpi-forum] make check is busted!

Tue Jan 31 16:01:01 CST 2012

I think the errors in the profiling chapters are benign: 

>Checking examples in ./chap-prof
>Opening file prof.tex
>ex256.c: In function 'PMPI_Example':
>ex256.c:11: warning: control reaches end of non-void function
>Error or warning detected in example on line 296 (prof.tex):
>ex296.c:7: error: 'FUNCTION' declared as function returning a function

This has to do with the FUNCTION() macro used in the example, being treated as a function itself.  I don't have the slightest clue how to 'fix' this.

>Example contains ... at line 325
>Error or warning detected in example on line 321 (prof.tex):
>ex321.c:6: error: expected identifier or '(' before '%' token

This one is probably a parsing issue, as it is not really code, but an example command line...

-Fab

Bronis R. de Supinski wrote on Tue, 31 Jan 2012 at 11:51:03

> 
> Hmm. That could explain why Fab found issues with the
> profiling chapter. I checked it in the past and I have
> not modified the examples so they should still work...
> 
> On Tue, 31 Jan 2012, Jeff Squyres wrote:
> 
>> Ok.
>> 
>> I should also note that some of the warnings that come up when you
> "make check" are dependent on your MPI implementation.  E.g., in the F90
> examples, if your "use mpi" module doesn't explicitly declare MPI functions,
> you may get warnings that they aren't declared.  But that's not a problem in
> the example code; it's a limitation of your MPI implementation.
>> 
>> 
>> 
>> On Jan 31, 2012, at 10:51 AM, William Gropp wrote:
>> 
>>> These should be managed by the chapter committees, and voted on a
> part of the chapter.  It would be counterproductive to vote on these
> individually.
>>> 
>>> Operationally, it is very difficult to mark up the text within the
>>> verbatim environment - frankly, the effort to do so is not worthwhile.
>>> 
>>> My recommendation is to make the chapter committees responsible for
> these.  Note that in some cases, the "fixes" are to the annotations, and do
> not change the text, and hence need no vote.
>>> 
>>> Bill
>>> 
>>> On Jan 31, 2012, at 9:28 AM, Jeff Squyres wrote:
>>> 
>>>> Bill --
>>>> 
>>>> If we make minor changes in examples in the text, do we need to
>>>> annotate them with "ticket 0" so that they can be ultimately voted on?
>>>> 
>>>> On Jan 31, 2012, at 10:16 AM, William Gropp wrote:
>>>> 
>>>>> Fab,
>>>>> 
>>>>> Yes, make check should build cleanly.  For example, the errors in
> coll.tex are Fortran 90 subroutines, listed as complete routines, that are
> missing the required "END" statement.  There are also "naked" examples
> that don't have the proper annotations; these should be added.
>>>>> 
>>>>> For those that don't know about the "make check" option - this makes
> sure that our examples don't have simple errors in them - it requires a small
> amount of work to add the necessary annotations (documented and
> examples exist throughout the source) and our experience has been that
> this catches a lot of errors that would otherwise be missed.  In looking over
> the make check output, I see quite a few of those.  Chapter authors should
> check their chapters.
>>>>> 
>>>>> Bill
>>>>> 
>>>>> On Jan 30, 2012, at 4:07 PM, Fab Tillier wrote:
>>>>> 
>>>>>> Hey Folks,
>>>>>> 
>>>>>> Is it reasonable for me to expect that building the examples in the
> standard wouldn't generate a bunch of errors?  Do we build the examples
> before publishing drafts, etc?
>>>>>> 
>>>>>> Just running 'make check' on the approved/MPI-3.0 branch gives me a
>>>>>> bunch of errors, in the following files: -
>>>>>> chap-one-side/one-side-2.tex - chap-context/context.tex -
>>>>>> chap-coll/coll.tex - chap-binding/binding-2.tex -
>>>>>> chap-topol/topol.tex - chap-prof/prof.tex
>>>>>> 
>>>>>> Can the chapter owners figure out how to either fix or suppress these
> errors?  It makes it really hard to find errors I may have introduced - the
> signal to noise ratio is really lousy.  Alternatively, is there a 'correct' way of
> running 'make check' that won't generate these?  I ran autoconf in
> mpicompilechk, then configure.  Should I be specifying extra parameters or
> anything?
>>>>>> 
>>>>>> Thanks,
>>>>>> -Fab
>>>>>> 
>>>>>> 
>>>>>> _______________________________________________
>>>>>> mpi-forum mailing list
>>>>>> mpi-forum at lists.mpi-forum.org
>>>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
>>>>> 
>>>>> William Gropp
>>>>> Director, Parallel Computing Institute
>>>>> Deputy Director for Research
>>>>> Institute for Advanced Computing Applications and Technologies
>>>>> Paul and Cynthia Saylor Professor of Computer Science
>>>>> University of Illinois Urbana-Champaign
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> _______________________________________________
>>>>> mpi-forum mailing list
>>>>> mpi-forum at lists.mpi-forum.org
>>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
>>>> 
>>>> 
>>>> --
>>>> Jeff Squyres
>>>> jsquyres at cisco.com
>>>> For corporate legal information go to:
>>>> http://www.cisco.com/web/about/doing_business/legal/cri/
>>>> 
>>>> 
>>>> _______________________________________________
>>>> mpi-forum mailing list
>>>> mpi-forum at lists.mpi-forum.org
>>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
>>> 
>>> William Gropp
>>> Director, Parallel Computing Institute
>>> Deputy Director for Research
>>> Institute for Advanced Computing Applications and Technologies
>>> Paul and Cynthia Saylor Professor of Computer Science
>>> University of Illinois Urbana-Champaign
>>> 
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> mpi-forum mailing list
>>> mpi-forum at lists.mpi-forum.org
>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
>> 
>> 
>> --
>> Jeff Squyres
>> jsquyres at cisco.com
>> For corporate legal information go to:
>> http://www.cisco.com/web/about/doing_business/legal/cri/
>> 
>> 
>> _______________________________________________
>> mpi-forum mailing list
>> mpi-forum at lists.mpi-forum.org
>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum
>> 
> _______________________________________________
> mpi-forum mailing list
> mpi-forum at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-forum