[petsc-dev] configure error on Edison

[Satish Balay]

Mark,

Thanks for the update. Sounds like all cmake (version) installs on
edison would benefit from this patch.

[and we should be wary of using --download-cmake on it]

Satish

On Mon, 9 Feb 2015, Mark Adams wrote:

> FYI, this why 2.8.11.2 worked and 2.8.12.2 failed.  Nothing to do with real
> versions of Cmake.
> 
> Glad to hear 2.8.11.2 seems to do the trick! I asked around, and it sounds
> like we manually commented out the -rdynamic from the platform cmake file
> (Platform/Linux.cmake) for that version in order to get it to work on
> Edison.
> 
> On Mon, Feb 9, 2015 at 5:40 PM, Mark Adams <mfadams at lbl.gov> wrote:
> 
> > FYI, this is what NERSC tells me. I tired cmake/3.0.0 but I got an error
> > from PETSc I think saying Metis needs v2.8.  (This should probably be
> > fixed.)  The CRAYPE_LINK_TYPE=dynamic did not seem to play nice with my
> > configuration ('--with-shared-libraries=0') but cmake/2.8.11.2 seems to
> > be working.
> >
> > Mark
> >
> > Thanks for the additional details - indeed, it is the -rdynamic flag being
> > used in the config tests that is causing problems. One way to to get around
> > that is setting CRAYPE_LINK_TYPE to dynamic (that way -rdynamic won't
> > fail), and the other is to use 2.8.11.2 (not 2.8.12.2 or 3.0.0) which does
> > not add -rdynamic on the link line when building config tests. Would either
> > of these workaround be ok for your use case?
> >
> > Many thanks,
> > Scott French
> > NERSC User Services
> >
> >
> > On Sun, Feb 8, 2015 at 1:37 PM, Mark Adams <mfadams at lbl.gov> wrote:
> >
> >> Damn, this was working a  week ago.  I've tried many things and am now
> >> just doing a 'make config' in metis and see this -rdynamic thing.  I
> >> will ask NERSC.
> >> Thanks,
> >> Mark
> >>
> >>
> >> On Sun, Feb 8, 2015 at 10:34 AM, Satish Balay <balay at mcs.anl.gov> wrote:
> >>
> >>> On Sat, 7 Feb 2015, Mark Adams wrote:
> >>>
> >>> > I am getting errors again with metis.  This has come up on Hopper but
> >>> this
> >>> > looks different.
> >>> > Mark
> >>> >
> >>>
> >>> >>>>>>>>>>>>
> >>> -- Check for working C compiler: /opt/cray/craype/2.2.1/bin/cc -- broken
> >>>
> >>>
> >>>   /opt/cray/craype/2.2.1/bin/cc -fast -no-ipo
> >>>   CMakeFiles/cmTryCompileExec1628815744.dir/testCCompiler.c.o -o
> >>>   cmTryCompileExec1628815744 -rdynamic
> >>>
> >>>   /usr/bin/ld: dynamic STT_GNU_IFUNC symbol `strcmp' with pointer
> >>> equality in
> >>>
> >>> `/usr/lib64/gcc/x86_64-suse-linux/4.3/../../../../lib64/libc.a(strcmp.o)'
> >>>   can not be used when making an executable; recompile with -fPIE and
> >>> relink
> >>>   with -pie
> >>>
> >>> <<<<<<<<<<<<
> >>>
> >>> I have no idea why cmake is using -rdynamic here. [or how to tell it not
> >>> to do that..]
> >>>
> >>> Satish
> >>>
> >>
> >>
> >
>