[petsc-dev] Possible error running C/C++ src/snes/examples/tutorials/ex19 with 1 MPI process

[Antonio Trande]

On 10/13/2016 04:13 PM, Satish Balay wrote:
> On Thu, 13 Oct 2016, Antonio Trande wrote:
> 
>>>>> Latest builds:
>>>>> http://copr-fe.cloud.fedoraproject.org/coprs/sagitter/petsc/build/464373/
>>>
>>>> I'll try and see if I can reproduce this.. [in a 32bit VM]..
>>>
>>> To start off - I looked at all the 5 32bit build logs in 464373 - and
>>> they don't have any C or OpeMPI tests crashing anymore.  I remember
>>> seeing these crashes in some of the previous logs. Any idea what fixed
>>> this?
>>
>> There are errors.
>> https://copr-be.cloud.fedoraproject.org/results/sagitter/petsc/fedora-rawhide-i386/00464373-petsc/build.log.gz
> 
> Not with C example. It says:
>>>
> C/C++ example src/snes/examples/tutorials/ex19 run successfully with 1 MPI process
> <<<
> 
> However you are probably refering to the message 'Error detected during compile or link!'. See below.

Yes, i can ignore 'Error detected during compile or link!'.
I will apply your patch.

> 
> 
>>> And wrt the fortran example [with non-mpi build] crashing - somehow
>>> the following appears to make a difference:
>>>
>>> FFLAGS=-specs=/usr/lib/rpm/redhat/redhat-hardened-cc1
> 
>>> Is it possible to remove this option from FFLAGS? [with 'FOPTFLAGS=']
>>
>> They are flags automatically set for "hardened" builds
>> (https://fedoraproject.org/wiki/Changes/Harden_All_Packages#Detailed_Harden_Flags_Description).
> 
> It doesn't mention fortran - and:
>>>>>>>>>>>>
>                            Non-hardening 	Hardening
> Compile for dynamic linking 	-fPIC 	        -fPIC
> <<<<<<<<<<
> 
> We already use -fPIC - so it shouldn't have made any difference - so a gfortran bug on i386?

I think no.
Same default flags are automatically set during RPM building; i will set
them manually.

>> New builds:
>>
>> Without "/usr/lib/rpm/redhat/redhat-hardened-cc1" flags:
>> http://copr-fe.cloud.fedoraproject.org/coprs/sagitter/petsc/build/464687/
>>
>> Same configuration for debugging:
>> http://copr-fe.cloud.fedoraproject.org/coprs/sagitter/petsc/build/464692/
>>
>Why a different debug build? Whats the difference? And what I should be
looking for? {in the debug build?}

In debug build aren't used default Fedora flags, only "-O0 -g".
'make test' uses valgrind.
Just to see if 'Error detected during compile or link!' warning was
still there.

>The CFLAGS are getting used once in <seq> - twice in <openmp> and
>three times in <mpich> builds.  Perhaps the FLAGS cleanup didn't fix
>all the builds.. [I won't cause problems - just make it a bit more
>nreadable..]

Probably happens because of built-in scripts that i use during
libpetsc-MPI builds.

I guess, we can consider all tests successfully passed.

-- 
---
Antonio Trande
mailto: sagitter 'at' fedoraproject 'dot' org
http://fedoraos.wordpress.com/
https://fedoraproject.org/wiki/User:Sagitter
GPG Key: 0x6CE6D08A
Check on https://keys.fedoraproject.org/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 473 bytes
Desc: OpenPGP digital signature
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20161013/e9d15e58/attachment.sig>