Path oddity in PNetCDF build/configure

[Carl Ponder]

Ok -- yes -- I got a clean build by removing the--with-mpi flag.

What about this --enable-large-single-req flag, though?
Does it require that I build OpenMPI specially, or use a specific kind 
of switch or FileSystem?
I only use 64-but systems these days...

Thanks,

                     Carl

------------------------------------------------------------------------
Subject: 	Re: Path oddity in PNetCDF build/configure
Date: 	Sun, 24 Oct 2021 21:27:44 +0000
From: 	Wei-Keng Liao <wkliao at northwestern.edu>
To: 	Carl Ponder <cponder at nvidia.com>
CC: 	Bruno Pagani <bruno.pagani at astrophysics.eu>, 
parallel-netcdf at lists.mcs.anl.gov <parallel-netcdf at lists.mcs.anl.gov>



*External email: Use caution opening links or attachments*


I checked the configure command line used.
While debugging this issue, I suggest you to remove
"--with-mpi --enable-large-single-req" from the configure command.

Option --with-mpi was designed to let users to provide the
path of MPI installation, e.g. --with-mpi=/usr/local

Use option --enable-large-single-req is when the underlying
MPI-IO supports reading and writing more than 2GiB in a
single call (i.e. MPI_File_read and MPI_File_write call.)

Wei-keng

> On Oct 24, 2021, at 4:07 PM, Carl Ponder <cponder at nvidia.com 
> <mailto:cponder at nvidia.com>> wrote:
>
>
> I attached theconfig.loghere, but it doesn't show the steps used in 
> thempiccchecking:
>
>     configure:4284: result: /usr/bin/sed -i''
>     *configure:4682: checking mpicc under yes*
>     configure:4689: result: no
>     configure:4772: error:
>     -----------------------------------------------------------------------
>          Specified MPI C compiler "mpicc" cannot be found under yes
>          PnetCDF requires a working MPI C compiler. Please specify the
>     location
>          of an MPI C compiler, either in the MPICC environment variable
>          (not CC variable) or through --with-mpi configure flag. Abort.
>     -----------------------------------------------------------------------
>
> The command-line I used was
>
>     ./configure --prefix=/usr/local --enable-shared --with-mpi
>     --enable-large-single-req
>
>
> ------------------------------------------------------------------------
> Subject: 	Re: Path oddity in PNetCDF build/configure
> Date:	Sun, 24 Oct 2021 19:22:06 +0000
> From:	Wei-Keng Liao<wkliao at northwestern.edu>
> To:	Carl Ponder<cponder at nvidia.com>
> CC:	Bruno 
> Pagani<bruno.pagani at astrophysics.eu>,parallel-netcdf at lists.mcs.anl.gov<parallel-netcdf at lists.mcs.anl.gov> 
>
>
>
>
> *External email: Use caution opening links or attachments*
>
>
> Please let me know the configure command used.
>
> Wei-keng
> ------------------------------------------------------------------------
>
> Subject: 	Re: Path oddity in PNetCDF build/configure
> Date: 	Sun, 24 Oct 2021 17:06:01 +0000
> From: 	Wei-Keng Liao<wkliao at northwestern.edu>
> To: 	Carl Ponder<cponder at nvidia.com>
> CC: 	parallel-netcdf at lists.mcs.anl.gov<parallel-netcdf at lists.mcs.anl.gov>
>
>
>
> *External email: Use caution opening links or attachments*
>
>
> Hi, Carl
>
> Please send me the file config.log.
>
> Wei-keng
>
> ------------------------------------------------------------------------
>> On Oct 24, 2021, at 1:29 PM, Carl Ponder <cponder at nvidia.com 
>> <mailto:cponder at nvidia.com>> wrote:
>>
>>
>> This is in a Dockerfile rather than my usual bash scripting.
>> Here's thePATHbeing printed during the build
>>
>>     Step 95/156 : RUN echo ${PATH}
>>      ---> Running in 8586a4aef68a
>>     /usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/local/bin:/usr/local/cuda/bin:/usr/local/pgi/Linux_x86_64/21.9/compilers/bin
>>
>> And thempicclocation tests correctly:
>>
>>     Step 95/155 : RUN which mpicc
>>      ---> Running in e2161d554278
>>     */usr/local/bin/mpicc*
>>
>> ------------------------------------------------------------------------
>> Subject:	Re: Path oddity in PNetCDF build/configure
>> Date:	Sun, 24 Oct 2021 16:32:42 +0400
>> From:	Bruno Pagani<bruno.pagani at astrophysics.eu>
>> To:	Carl Ponder<cponder at nvidia.com>, Wei-keng 
>> Liao<wkliao at eecs.northwestern.edu>
>> CC:	parallel-netcdf at lists.mcs.anl.gov
>>
>>
>>
>> *External email: Use caution opening links or attachments*
>>
>>
>> How do you set PATH? Did you quote it? Seems somehow your path end up 
>> being yes, likely due to expansion of its previous value.
>>
>>
>> ------------------------------------------------------------------------
>> Le 24 octobre 2021 16:30:03 GMT+04:00, Carl 
>> Ponder<cponder at nvidia.com>a écrit :
>>
>>     When I build PNetCDF withMPICC=mpiccandmpiccon my$PATH, I see
>>     this during theconfigurephase:
>>
>>         ENV PATH=${PATH}:/usr/local/bin
>>         ENV MPICC mpicc
>>         ...
>>         checking for sed handling option -i ... /usr/bin/sed -i''
>>         *checking mpicc under yes... no*
>>         configure: error:
>>         -----------------------------------------------------------------------
>>         Specified MPI C compiler "mpicc" cannot be found under yes
>>         PnetCDF requires a working MPI C compiler. Please specify the
>>         location
>>         of an MPI C compiler, either in the MPICC environment variable
>>         (not CC variable) or through --with-mpi configure flag. Abort.
>>         -----------------------------------------------------------------------
>>
>>     I can work around it by making the path absolute
>>
>>         ENV MPICC /usr/local/bin/mpicc
>>         ...
>>         checking for sed handling option -i ... /usr/bin/sed -i''
>>         *checking whether /usr/local/bin/mpicc exists and is
>>         executable... yes*
>>
>>     But, based on the first message, I'm thinking there's an
>>     incorrect path-search during theconfigurephase.
>>     This is with PNetCDF 1.12.2.
>>
>
> <config.log>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/parallel-netcdf/attachments/20211024/766d8bad/attachment-0001.html>