Well - it's BG/Q... so ultimately it is using some weird compiler - but I'm
using a GCC "wrapper" for them so that it looks like GCC on the outside.
 It's all very tough to explain...

BUT - the admins of the machine pointed me to the following:

http://www.gnu.org/software/libtool/manual/html_node/Configure-notes.html

And in the last bullet there it says that you can set some autoconf
variables at configure time to restrict the paths libtool searches in.

BUT - I really don't understand.  Does this mean when configuring libMesh -
or when configuring the build of libtool - does anyone know how or when I
would set these variables?

Derek





On Tue, Dec 10, 2013 at 10:11 AM, Paul T. Bauman <ptbau...@gmail.com> wrote:

> This is strange. What compilers are you using? This almost looks to me
> like they did a system update, but didn't update the system gcc install.
>
> In attempt to just and get you running ASAP (apologies if you already
> tried this), that looks like a core system path and that will probably be
> inside the libtool script (it's just a bash script). As a quick, ugly ugly
> hack, can you search through the libtool script for V1R2M0 and replace with
> the new path name? Note the libtool script gets generated after configure,
> so if you rerun configure, you're changes will get wiped out.
>
>
> On Tue, Dec 10, 2013 at 10:38 AM, Derek Gaston <fried...@gmail.com> wrote:
>
>> So - it turns out that this happens with the system libtool as well (just
>> a little later on in the compile).
>>
>> Any ideas for tracking this down?
>>
>> This is kind of a big deal as I have a big code review coming up next
>> week and I need this to work...
>>
>> Derek
>>
>>
>>
>> On Mon, Dec 9, 2013 at 9:51 PM, Derek Gaston <fried...@gmail.com> wrote:
>>
>>> The offending command:
>>>
>>> ../../libtool --verbose --tag=CXX   --mode=link mpicxx -O2
>>> -felide-constructors -funroll-loops -fstrict-aliasing -std=c++0x
>>> -Wdisabled-optimization   -fopenmp  -all-static  -o libopt.la
>>> libopt_la-tetgen.lo libopt_la-predicates.lo
>>>
>>> (executed in contirb/tetgen in the build directory)
>>>
>>> I did a bit of digging and the correct file appears to be:
>>>
>>>
>>> /bgsys/drivers/toolchain/V1R2M0-efix23/gnu-linux/powerpc64-bgq-linux/lib/libstdc++.la
>>>
>>> Notice the "-efix23" that is not present in the directory our libtool is
>>> looking for.  Where does libtool get this directory?  This could actually
>>> be a misconfiguration on the machine that the admins can fix... but I don't
>>> know how to hunt down the issue.
>>>
>>> Note that libtool --verbose doesn't work here - because it's actually
>>> trying to build the command it's going to run when grep actually exits - so
>>> I can't see exactly what it's trying to do...
>>>
>>> Derek
>>>
>>>
>>>
>>>
>>>
>>> On Mon, Dec 9, 2013 at 9:29 PM, Derek Gaston <fried...@gmail.com> wrote:
>>>
>>>> Running bootstrap doesn't fix it.
>>>>
>>>> Neither does running "autoreconf -iv --force"
>>>>
>>>> Only thing that fixes it so far is soft-linking the system libtool...
>>>>
>>>>
>>>> On Mon, Dec 9, 2013 at 9:16 PM, Derek Gaston <fried...@gmail.com>wrote:
>>>>
>>>>> It's a problem when building tetgen (well - it shows up there first):
>>>>>
>>>>>
>>>>>    CXXLD    libopt.la
>>>>>
>>>>> /bin/grep:
>>>>> /bgsys/drivers/toolchain/V1R2M0/gnu-linux/powerpc64-bgq-linux/lib/libstdc++.la:
>>>>> No such file or directory
>>>>>
>>>>> /bin/sed: can't read
>>>>> /bgsys/drivers/toolchain/V1R2M0/gnu-linux/powerpc64-bgq-linux/lib/libstdc++.la:
>>>>> No such file or directory
>>>>>
>>>>> libtool: link:
>>>>> `/bgsys/drivers/toolchain/V1R2M0/gnu-linux/powerpc64-bgq-linux/lib/libstdc++.la'
>>>>> is not a valid libtool archive
>>>>>
>>>>> make[2]: *** [libopt.la] Error 1
>>>>>
>>>>> make[2]: Leaving directory
>>>>> `/gpfs/mira-home/gastdr/projects/libmesh/build/contrib/tetgen'
>>>>>
>>>>> The weird thing is that /bgsys/drivers/toolchain/V1R2M0 doesn't
>>>>> exist... I have no idea no idea how it's picking it up... but the system
>>>>> libtool doesn't have that issue...
>>>>>
>>>>> Derek
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On Mon, Dec 9, 2013 at 2:52 PM, Kirk, Benjamin (JSC-EG311) <
>>>>> benjamin.k...@nasa.gov> wrote:
>>>>>
>>>>>>
>>>>>> On Dec 9, 2013, at 3:33 PM, John Peterson <jwpeter...@gmail.com>
>>>>>>  wrote:
>>>>>>
>>>>>> > On Mon, Dec 9, 2013 at 2:19 PM, Derek Gaston <fried...@gmail.com>
>>>>>> wrote:
>>>>>> >> So - I'm back at trying to get everything running on BG/Q (Mira) at
>>>>>> >> Argonne... and I hit a snag where our supplied libtool is doing
>>>>>> something
>>>>>> >> bad - but if I just move it out of the way and set a soft-link to
>>>>>> the system
>>>>>> >> libtool then everything is working fine.
>>>>>> >>
>>>>>> >> Is there a way to tell our build system not to use our included
>>>>>> libtool at
>>>>>> >> all?
>>>>>> >
>>>>>> > Is the error during building libmesh?  What is the error?
>>>>>>
>>>>>> I'm interested in the answer as well, but...
>>>>>>
>>>>>> boostrap has some simple logic in there to look for a preferred
>>>>>> version of libtool and build ours.  But at the end of the day, all
>>>>>> bootstrap does is run
>>>>>>
>>>>>> $ autoreconf -iv --force
>>>>>>
>>>>>> so if your path is set up such that
>>>>>>
>>>>>> 1) autoreconf
>>>>>> 2) automake
>>>>>> 3) libtool
>>>>>>
>>>>>> point to the versions you desire, just run
>>>>>>
>>>>>> $  autoreconf -iv --force
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
>>
>> ------------------------------------------------------------------------------
>> Rapidly troubleshoot problems before they affect your business. Most IT
>> organizations don't have a clear picture of how application performance
>> affects their revenue. With AppDynamics, you get 100% visibility into your
>> Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics
>> Pro!
>>
>> http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
>> _______________________________________________
>> Libmesh-devel mailing list
>> Libmesh-devel@lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/libmesh-devel
>>
>>
>
------------------------------------------------------------------------------
Rapidly troubleshoot problems before they affect your business. Most IT 
organizations don't have a clear picture of how application performance 
affects their revenue. With AppDynamics, you get 100% visibility into your 
Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics Pro!
http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel

Reply via email to