[Darshan-users] undefined reference to __wrap_printf

Phil Carns carns at mcs.anl.gov
Fri Jul 21 15:05:38 CDT 2017


It might be helpful to compare the mpich configure/compile lines between 
your two configurations too, in case there is some difference in mpich 
libraries that is triggering an issue.

thanks,
-Phil

On 07/20/2017 05:35 PM, Shane Snyder wrote:
>
> On 07/20/2017 04:19 PM, Shane Snyder wrote:
>> Hi Teng,
>>
>> First of all, I actually just got darshan 3.1.4 to properly generate 
>> a log when using static linking. I'm not sure what was leading to the 
>> problem I had earlier, but after reconfiguring and rebuilding the 
>> library, it works for me now. I also confirmed the nightly tests were 
>> working correctly -- the log files were going to another spot that I 
>> wasn't expecting, so missed them the first time around. You may want 
>> to try something similar -- just rebuilding everything from scratch 
>> and trying again. I've also run into this problem in the past 
>> whenever I was accidentally pointing to an mpich build that defaults 
>> to using shared libraries rather than the static ones, so that could 
>> be another thing to check.
>>
>> As for the error you are getting with Darshan 3.1.3, I've actually 
>> never seen someone hit that before. That error would be due to the 
>> libz compression functions (deflate) returning some sort of error. 
>> Not exactly sure what could cause that, but is the application 
>> opening a really large amount of files or something like that? I 
>> suppose it's possible the compression routines could be running out 
>> of memory if there is too much log data they are trying to compress, 
>> but have never seen that happen before. If you aren't already, it 
>> might be worth trying a test program that does very basic I/O just to 
>> confirm that we can get Darshan working. One that we frequently use 
>> is mpi-io-test -- there is a version of this code in the darshan 
>> source @ darshan-test/regression/test-cases/src. You could try just 
>> building that and running it (it doesn't take any command line 
>> parameters) to see if that works?
>
> Oh, actually I just noticed in your emails that you were using 
> mpi-io-test? That is pretty bizarre...
>
> What system are you running this on? A cluster? Your own personal 
> machine? If possible, you might try running somewhere else so we can 
> confirm the issue is some problem with the environment on the system 
> you have been using... I can't think of why the deflate routines would 
> fail when running something simple like mpi-io-test. But if needed, we 
> could probably get you a patch or something that tries to give us more 
> details on the libz error you are seeing.
>
> --Shane
>
>>
>> Thanks,
>> --Shane
>>
>> On 07/20/2017 03:52 PM, Teng Wang wrote:
>>> Hi Shane,
>>>
>>> Thanks for your help on this. I tried v3.1.3 and compiled successfully.
>>> But when I ran the application, there was no *.log generated under the
>>> log directory, and I also got a warning from the application:
>>>
>>> darshan library warning: unable to compress job data
>>>
>>>
>>> May I know how to fix this?
>>>
>>> Thanks,
>>> Teng
>>>
>>> On Thu, Jul 20, 2017 at 11:51 AM, Shane Snyder <ssnyder at mcs.anl.gov 
>>> <mailto:ssnyder at mcs.anl.gov>> wrote:
>>>
>>>     Hi Teng,
>>>
>>>     Thanks for reporting this. This is a really strange issue, but I
>>>     can actually reproduce it, too. I'm going to keep digging and
>>>     see if I can find the problem. Will keep you updated.
>>>
>>>     The strange thing is that I was able to use version 3.1.4
>>>     successfully right before making the release available (i.e.,
>>>     the log @
>>>     darshan-test/example-output/mpi-io-test-x86_64-3.1.4.darshan was
>>>     generated using version 3.1.4 with static linking). Also strange
>>>     is that our nightly testing is not catching this issue...
>>>
>>>     FYI, version 3.1.3 and earlier work fine for me if you really
>>>     need a quick resolution -- version 3.1.4 just fixed a couple of
>>>     non-critical bugs in some darshan log parsing utilities, so
>>>     there shouldn't be much functionally different between the versions.
>>>
>>>     --Shane
>>>
>>>
>>>     On 07/20/2017 12:43 PM, Teng Wang wrote:
>>>>
>>>>     Hi,
>>>>
>>>>
>>>>     I'm having trouble when trying to statically link
>>>>
>>>>     Darshan with mpich3.3 applications following the instructions
>>>>
>>>>     here. Could you give me any suggestion on how to fix it?
>>>>
>>>>
>>>>     http://www.mcs.anl.gov/research/projects/darshan/docs/darshan-runtime.html
>>>>     <http://www.mcs.anl.gov/research/projects/darshan/docs/darshan-runtime.html>
>>>>
>>>>
>>>>     Here is my steps:
>>>>
>>>>     1. Compile and install darshan 3.1.4
>>>>
>>>>     tar -xvzf darshan-3.1.4.tar.gz
>>>>
>>>>     cd darshan-3.1.4/darshan-runtime
>>>>
>>>>     ./configure --with-mem-align=8 --with-log-path=<log directory>
>>>>     --prefix=<install directory> --with-jobid-env=PBS_JOBID
>>>>     --disable-cuserid CC=mpicc
>>>>
>>>>     make
>>>>
>>>>     make install
>>>>
>>>>
>>>>     2. Create log directory, which successfully create the directory
>>>>
>>>>     darshan-mk-log-dirs.pl <http://darshan-mk-log-dirs.pl/>
>>>>
>>>>
>>>>     3. Generate wrapper
>>>>
>>>>     darshan-gen-cc.pl <http://darshan-gen-cc.pl/> `which mpicc`
>>>>     --output mpicc.darshan
>>>>
>>>>
>>>>     4. Compile application (mpi-io-test.c) using mpicc.darshan
>>>>
>>>>     mpicc.darshan mpi-io-test.c -o mpi-io-test
>>>>
>>>>
>>>>     After step 4, the following issue happened:
>>>>
>>>>     initthread.c:(.text+0xc59d): undefined reference to
>>>>     `__wrap_fprintf'
>>>>
>>>>     initthread.c:(.text+0xc5bd): undefined reference to `__wrap_fputs'
>>>>
>>>>     /global/homes/t/user/software_install/mpich3.3a-static/lib/libmpi.a(lib_libmpi_la-finalize.o):
>>>>     In function `MPIR_Add_finalize':
>>>>
>>>>     finalize.c:(.text+0x226e): undefined reference to `__wrap_fputs'
>>>>
>>>>     finalize.c:(.text+0x227a): undefined reference to `__wrap_fflush'
>>>>
>>>>     finalize.c:(.text+0x2286): undefined reference to `__wrap_fflush'
>>>>
>>>>     /global/homes/t/user/software_install/mpich3.3a-static/lib/libmpi.a(lib_libmpi_la-util.o):
>>>>     In function `MPIDI_OFI_control_handler':
>>>>
>>>>     util.c:(.text+0x17bc4): undefined reference to `__wrap_fprintf'
>>>>
>>>>     /global/homes/t/user/software_install/mpich3.3a-static/lib/libmpi.a(lib_libmpi_la-ch4_globals.o):
>>>>     In function `MPID_Abort':
>>>>
>>>>     ch4_globals.c:(.text+0x38b): undefined reference to `__wrap_fputs'
>>>>
>>>>     ch4_globals.c:(.text+0x397): undefined reference to `__wrap_fflush'
>>>>
>>>>     ch4_globals.c:(.text+0x3a3): undefined reference to `__wrap_fflush'
>>>>
>>>>     /tmp/ccyntLoA.o: In function `main':
>>>>
>>>>     mpi-io-test.c:(.text+0x5ae): undefined reference to `__wrap_printf'
>>>>
>>>>     mpi-io-test.c:(.text+0x819): undefined reference to `__wrap_printf'
>>>>
>>>>     /global/homes/t/user/software_install/mpich3.3a-static/lib/libmpi.a(lib_libmpi_la-contextid.o):
>>>>     In function `MPIR_Free_contextid':
>>>>
>>>>     contextid.c:(.text+0x1c3f): undefined reference to `__wrap_fputs'
>>>>
>>>>     contextid.c:(.text+0x1c4b): undefined reference to `__wrap_fflush'
>>>>
>>>>     contextid.c:(.text+0x1c57): undefined reference to `__wrap_fflush'
>>>>
>>>>
>>>>
>>>>     mpicc.darshan -show mpi-io-test.c -o mpi-io-test gave:
>>>>
>>>>
>>>>     gcc -L/usr/lib64/slurmpmi
>>>>     -L/global/common/cori/software/libfabric/1.4.1/gnu/lib
>>>>     mpi-io-test.c -o mpi-io-test
>>>>     -L/global/homes/t/user/software_install/darshan/lib -ldarshan
>>>>     -lz
>>>>     -Wl,@/global/homes/t/user/software_install/darshan/share/ld-opts/darshan-base-ld-opts
>>>>     -I/global/homes/t/user/software_install/mpich3.3a-static/include
>>>>     -L/global/homes/t/user/software_install/mpich3.3a-static/lib
>>>>     -lmpi -lpmi -lpmi -lpthread -lfabric -lrt -lpmi
>>>>     -L/global/homes/t/user/software_install/darshan/lib
>>>>     -Wl,--start-group -ldarshan -ldarshan-stubs -Wl,--end-group -lz
>>>>     -lrt -lpthread
>>>>
>>>>
>>>>     The strange thing is when I do
>>>>
>>>>     nm libdarshan.a, it did show the functions (e.g. __wrap_printf)
>>>>     were linked to
>>>>
>>>>     libdarshan.a.
>>>>
>>>>
>>>>     *0000000000002bf0 T __wrap__IO_getc*
>>>>
>>>>     *0000000000002d80 T __wrap__IO_putc*
>>>>
>>>>     *00000000000030b0 T __wrap___isoc99_fscanf*
>>>>
>>>>     *0000000000001990 T __wrap_fclose*
>>>>
>>>>     *00000000000011d0 T __wrap_fdopen*
>>>>
>>>>     *0000000000001810 T __wrap_fflush*
>>>>
>>>>     *0000000000002a60 T __wrap_fgetc*
>>>>
>>>>     *0000000000003690 T __wrap_fgets*
>>>>
>>>>     *0000000000000db0 T __wrap_fopen*
>>>>
>>>>     *0000000000000fc0 T __wrap_fopen64*
>>>>
>>>>     *00000000000026b0 T __wrap_fprintf*
>>>>
>>>>     *0000000000001c80 T __wrap_fputc*
>>>>
>>>>     *0000000000001fa0 T __wrap_fputs*
>>>>
>>>>     *00000000000028d0 T __wrap_fread*
>>>>
>>>>     *00000000000013d0 T __wrap_freopen*
>>>>
>>>>     *00000000000015f0 T __wrap_freopen64*
>>>>
>>>>     *00000000000032d0 T __wrap_fscanf*
>>>>
>>>>     *0000000000003920 T __wrap_fseek*
>>>>
>>>>     *0000000000003a30 T __wrap_fseeko*
>>>>
>>>>     *0000000000003b40 T __wrap_fseeko64*
>>>>
>>>>     *0000000000003c50 T __wrap_fsetpos*
>>>>
>>>>
>>>>     Thanks,
>>>>
>>>>     Teng
>>>>
>>>>
>>>>
>>>>     _______________________________________________
>>>>     Darshan-users mailing list
>>>>     Darshan-users at lists.mcs.anl.gov
>>>>     <mailto:Darshan-users at lists.mcs.anl.gov>
>>>>     https://lists.mcs.anl.gov/mailman/listinfo/darshan-users
>>>>     <https://lists.mcs.anl.gov/mailman/listinfo/darshan-users>
>>>
>>>
>>>     _______________________________________________
>>>     Darshan-users mailing list
>>>     Darshan-users at lists.mcs.anl.gov
>>>     <mailto:Darshan-users at lists.mcs.anl.gov>
>>>     https://lists.mcs.anl.gov/mailman/listinfo/darshan-users
>>>     <https://lists.mcs.anl.gov/mailman/listinfo/darshan-users>
>>>
>>>
>>
>>
>>
>> _______________________________________________
>> Darshan-users mailing list
>> Darshan-users at lists.mcs.anl.gov
>> https://lists.mcs.anl.gov/mailman/listinfo/darshan-users
>
>
>
> _______________________________________________
> Darshan-users mailing list
> Darshan-users at lists.mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/darshan-users


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/darshan-users/attachments/20170721/be9d371a/attachment-0001.html>


More information about the Darshan-users mailing list