[Pkg-openmpi-maintainers] fortran openmpi on mipsel

[Drew Parsons]

On Mon, 2016-04-04 at 14:56 +0100, James Cowgill wrote:
> Hi,
> 
> On Mon, 2016-04-04 at 11:52 +0800, Drew Parsons wrote:
> > 
> > Dear mipsel and openmpi developers,
> > 
> > in the build logs for petsc something has gone wrong while checking
> > the
> > fortran openmpi installation on mipsel.
> > 
> > https://buildd.debian.org/status/fetch.php?pkg=petsc&arch=mipsel&ve
> > r=3.6.3.dfsg2-1&stamp=1459708874
> This is probably caused by bug #818909 in openmpi (ie openmpi being
> totally broken on mips*). It doesn't look like there's been any
> progress on it though.
> 


Thanks for the feedback James.  That bug reports mpicc rather than
fortran. But you're right, even though my petsc build stops on the
fortran check, further up the log ("TESTING: checkCCompiler") I do see
a Segmentation fault from mpicc:

        Checking for program /usr/bin/mpicc...found
                    Defined make macro "CC" to "mpicc"
                    Pushing language C
                              All intermediate test results are stored in /tmp/petsc-QnNaJd/config.setCompilers
        Executing: mpicc -c -o /tmp/petsc-QnNaJd/config.setCompilers/conftest.o -I/tmp/petsc-QnNaJd/config.setCompilers -g -O2 -fstack-protector-strong -Wformat -Werror=format-security -Wdate-time -D_FORTIFY_SOURCE=2 /tmp/petsc-QnNaJd/config.setCompilers/conftest.c 
        Possible ERROR while running compiler: exit code 35584
        stderr:
        [mipsel-manda-02:10115] *** Process received signal ***
        [mipsel-manda-02:10115] Signal: Segmentation fault (11)
        [mipsel-manda-02:10115] Signal code: Address not mapped (1)
        [mipsel-manda-02:10115] Failing at address: 0x1aadf988
        [mipsel-manda-02:10115] *** End of error message ***
        Segmentation fault

Funny that the petsc build script did not stop at this point.

I'll place a blocking link on my bug.

Drew