sieve-dev building petsc-dev on IBM AIX
Shi Jin
jinzishuai at gmail.com
Tue Jul 8 15:39:51 CDT 2008
Hi,
I figured out that the solution on AIX to have two quotes together is to do
the following to the $PETSC_DIR/conf/variables, at line 18:
-D__SDIR__=" '\"${LOCDIR}\"' "
This way, I would have
-D__SDIR__= '"src/dm/mesh/"' during the make on AIX.
Howver, the same change does not work on Linux. So I guess this have to do
dealt with separately in the configuration process.
I guess this somehow solved the first problem. I still don't know what is
wrong with the error:
"sieve/Mesh.hh", line 1195.30: 1540-1110 (S) The referenced type
"ALE::IBundle" contains a circular reference back to "".
Thanks.
Shi
On Sat, Jun 28, 2008 at 6:04 PM, Shi Jin <jinzishuai at gmail.com> wrote:
> Hi,
>
> I've noticed that changes are made to the latest petsc-dev source so that
> the compilation has a quote like
> mpCC_r -q64 -o mesh.o -c -qrtti=dyna -O -+ -Isieve
> -I/scratch_ibm/seki/software/src/petsc-dev/externalpackages/Boost/
> -I/home/seki/software/src/petsc-dev
> -I/home/seki/software/src/petsc-dev/aix5.3.0.0-cxx-opt/include
> -I/home/seki/software/src/petsc-dev/include
> -I/scratch_ibm/seki/software/src/petsc-dev/src/dm/mesh/sieve
> -I/usr/lpp/ppe.poe/include
> -I/scratch_ibm/seki/software/src/petsc-dev/externalpackages/Boost/*-D__SDIR__="src/dm/mesh/"
> * mesh.c
>
> However, this is still not correct since one pair of single quotes are
> missing. It should be * -D__SDIR__='"src/dm/mesh/"'* . Could you please
> look into this furthermore to correct the AIX behavior?
>
> Even this is corrected as above, I still get three error messages (produced
> with the -w option to suppress warnings):
>
> "sieve/Mesh.hh", line 1181.30: 1540-1110 (S) The referenced type
> "ALE::IBundle" contains a circular reference back to "".
> "mesh.c", line 1622.39: 1540-0130 (S)
> "ALE::Generator<ALE::IMesh>::refineMeshV" is not declared.
> "mesh.c", line 1643.39: 1540-0130 (S)
> "ALE::Generator<ALE::IMesh>::refineMeshV" is not declared.
>
> I would appreciate it if you could take a look at this problem after
> solving the first.
> Thank you very much.
>
> Shi
> On Wed, Jun 18, 2008 at 3:48 PM, Shi Jin <jinzishuai at gmail.com> wrote:
>
>> I did the test on src/dm/mesh/
>> On my linux system, the mesh.c is built by (got it by running make mesh.o)
>>
>> g++ -o newmesh.c -E -Wall -Wwrite-strings -Wno-strict-aliasing -g
>> -Isieve -I/mnt/UofA/sekidata/petsc-dev/externalpackages/Boost/
>> -I/home/seki/UofA/petsc-dev/linux-gnu-cxx-debug/include
>> -I/home/seki/UofA/petsc-dev/linux-gnu-cxx-debug/include
>> -I/home/seki/UofA/petsc-dev
>> -I/home/seki/UofA/petsc-dev/linux-gnu-cxx-debug/include
>> -I/home/seki/UofA/petsc-dev/include
>> -I/home/seki/UofA/petsc-dev/linux-gnu-cxx-debug/include
>> -I/mnt/UofA/sekidata/petsc-dev/src/dm/mesh/sieve -I/usr/local/include
>> -I/home/seki/UofA/petsc-dev/linux-gnu-cxx-debug/include
>> -I/home/seki/UofA/petsc-dev/linux-gnu-cxx-debug/include
>> -I/mnt/UofA/sekidata/petsc-dev/externalpackages/Boost/ *-D__SDIR__='"src/dm/mesh/"'
>> * mesh.c
>> This one works fine.
>>
>> On the AIX system,, it is done by
>> mpCC_r -q64 -o mesh.o -c -qrtti=dyna -O -+ -Isieve
>> -I/scratch_ibm/seki/software/src/petsc-dev/externalpackages/Boost/
>> -I/home/seki/software/src/petsc-dev
>> -I/home/seki/software/src/petsc-dev/aix5.3.0.0-cxx-opt/include
>> -I/home/seki/software/src/petsc-dev/include
>> -I/scratch_ibm/seki/software/src/petsc-dev/src/dm/mesh/sieve
>> -I/usr/lpp/ppe.poe/include
>> -I/scratch_ibm/seki/software/src/petsc-dev/externalpackages/Boost/ *
>> -D__SDIR__=src/dm/mesh/* mesh.c
>> It gives lots of (S) error messages
>> After putting quotes to the *-D__SDIR__='"src/dm/mesh/"'* as in the
>> first case, those errors are gone.
>>
>> I hope this identifies the problem. I guess modification can be made to
>> some global makefile variables so that it is effective everywhere.
>> Please let me know if this is helpful.
>> Thanks.
>>
>> Shi
>>
>>
>> On Wed, Jun 18, 2008 at 11:45 AM, Matthew Knepley <knepley at gmail.com>
>> wrote:
>>
>>> Please send me the compile line ofr this file.
>>>
>>> Thanks,
>>>
>>> Matt
>>>
>>> On Wed, Jun 18, 2008 at 12:42 PM, Shi Jin <jinzishuai at gmail.com> wrote:
>>> > It looks like the problem is on the macro variable __SDIR__
>>> > After preprocessing, it shows up as a string without quotes, such as
>>> > src/dm/mesh
>>> > I think the solution is to put quotes to it.
>>> >
>>> > Shi
>>> > On Wed, Jun 18, 2008 at 11:24 AM, Shi Jin <jinzishuai at gmail.com>
>>> wrote:
>>> >>
>>> >> Hi,
>>> >>
>>> >> After looking at the problem in more details, I realized that the
>>> offsetof
>>> >> POD problem is not the critical one.
>>> >> It is just a warning.
>>> >> However, I had several more serious errors, which are all related to
>>> >> macros defined in petsc.
>>> >> I have found out that CHKERRQ, SETERRQ, SETERRQ1 and PetscHeaderCreate
>>> are
>>> >> all giving severe error messages like
>>> >> "mesh.c", line 414.10: 1540-0063 (S) The text "," is unexpected.
>>> >> "mesh.c", line 466.10: 1540-0063 (S) The text ")" is unexpected.
>>> >>
>>> >> I think they are caused by the macros but not sure exactly what the
>>> >> problem is and how to fix it.
>>> >> Shi
>>> >>
>>> >> On Tue, Jun 17, 2008 at 7:43 PM, Matthew Knepley <knepley at gmail.com>
>>> >> wrote:
>>> >>>
>>> >>> On Tue, Jun 17, 2008 at 7:40 PM, Shi Jin <jinzishuai at gmail.com>
>>> wrote:
>>> >>> > Hi,
>>> >>> >
>>> >>> > I noticed that changes have been made to make petsc-dev compilable
>>> on
>>> >>> > AIX
>>> >>> > systems with the xlc compilers. Therefore, I re-visited the same
>>> issue
>>> >>> > again with the latest petsc-dev code on an IBM P5 machine running
>>> AIX
>>> >>> > 5.3.0.50.
>>> >>> > I tried to build petsc-dev with
>>> >>> > PETSC_DIR=`pwd` PETSC_ARCH="aix5.3.0.0-cxx-opt" config/configure.py
>>> >>> > --with-debugging=no --with-cxx="mpCC_r -q64" --with-fortran=0
>>> >>> > --with-shared=1 --with-clanguage=C++ --with-sieve=1
>>> >>> > --download-boost=1
>>> >>> > PETSC_DIR=`pwd` PETSC_ARCH="aix5.3.0.0-cxx-opt" make
>>> >>> >
>>> >>> > The configuration went perfectly, without any complaint.
>>> >>> > However, I got lots of messages during the make:
>>> >>> > "../../sieve/Sifter.hh", line 560.180: 1540-1281 (W) "offsetof"
>>> cannot
>>> >>> > be
>>> >>> > applied to "struct Arrow<int,int,int>". It is not a POD (plain old
>>> >>> > data)
>>> >>> > type.
>>> >>> >
>>> >>> > Other than this problem, the rest seems fine. I got the same
>>> offsetof
>>> >>> > POD
>>> >>> > problem when compiling my code with the newly built petsc-dev
>>> library.
>>> >>> > I am
>>> >>> > very encouraged by with since last time we saw lots of different
>>> error
>>> >>> > messages.
>>> >>> >
>>> >>> > So I am wondering what we can do to solve the offsetof POD problem.
>>> We
>>> >>> > definitely did not see this with g++. . My version is
>>> >>> > seki at cortex conf $ mpCC_r -qversion
>>> >>> > IBM XL C/C++ Enterprise Edition V8.0 for AIX
>>> >>> > Version: 08.00.0000.0013
>>> >>>
>>> >>> I saw this as well. I tried to google the error message, but found
>>> >>> nothing to help.
>>> >>> Evidently Boost and MySQL also get this warning but it things still
>>> run.
>>> >>>
>>> >>> Matt
>>> >>>
>>> >>> > Thank you very much.
>>> >>> > Shi Jin
>>> >>> > On Tue, Feb 26, 2008 at 8:45 AM, Matthew Knepley <
>>> knepley at gmail.com>
>>> >>> > wrote:
>>> >>> >>
>>> >>> >> I have been trying to make sense of these compile errors, but only
>>> >>> >> some
>>> >>> >> of them are valid I believe. The rest seem to be bugs in the C++
>>> >>> >> compiler.
>>> >>> >> Are you sure you have the latest version? I suggest putting g++ on
>>> >>> >> this
>>> >>> >> machine first and building a valid copy. Then we can address the
>>> >>> >> native
>>> >>> >> compiler issues.
>>> >>> >>
>>> >>> >> Thanks,
>>> >>> >>
>>> >>> >> Matt
>>> >>> >>
>>> >>> >
>>> >>>
>>> >>>
>>> >>>
>>> >>> --
>>> >>> What most experimenters take for granted before they begin their
>>> >>> experiments is infinitely more interesting than any results to which
>>> >>> their experiments lead.
>>> >>> -- Norbert Wiener
>>> >>
>>> >>
>>> >>
>>> >> --
>>> >> Sincerely,
>>> >> Shi Jin, Ph.D.
>>> >> http://www.ualberta.ca/~sjin1/ <http://www.ualberta.ca/%7Esjin1/>
>>> >
>>> >
>>> > --
>>> > Sincerely,
>>> > Shi Jin, Ph.D.
>>> > http://www.ualberta.ca/~sjin1/ <http://www.ualberta.ca/%7Esjin1/>
>>>
>>>
>>>
>>> --
>>> What most experimenters take for granted before they begin their
>>> experiments is infinitely more interesting than any results to which
>>> their experiments lead.
>>> -- Norbert Wiener
>>>
>>
>>
>>
>> --
>> Sincerely,
>> Shi Jin, Ph.D.
>> http://www.ualberta.ca/~sjin1/ <http://www.ualberta.ca/%7Esjin1/>
>>
>
>
>
> --
> Sincerely,
> Shi Jin, Ph.D.
> http://www.ualberta.ca/~sjin1/ <http://www.ualberta.ca/%7Esjin1/>
>
--
Sincerely,
Shi Jin, Ph.D.
http://www.ualberta.ca/~sjin1/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mcs.anl.gov/mailman/private/sieve-dev/attachments/20080708/8eb97c2d/attachment.htm>
More information about the sieve-dev
mailing list