Unable to build HDF5 with pgi compilers

Hdf-forum archives
1

Hello,
         I already have an existing installation of HDF5 1.8.17 compiled
with gfortran. Now for another project on the same machine I need to
compile with PGI compilers i.e. pgcc , pgc++and pgf90. The installation
directory will be separate as well i.e. /opt/local/lib and
/opt/local/include.

This is how I try to build hdf5

CPPFLAGS=/opt/local/include CXXFLAGS=/opt//local/include
LDFLAGS=/opt//local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure
--prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx.

and when I try doing this I get this error message -

checking for a BSD-compatible install... /usr/bin/install -c
checking whether build environment is sane... yes
checking for a thread-safe mkdir -p... /bin/mkdir -p
checking for gawk... no
checking for mawk... mawk
checking whether make sets $(MAKE)... yes
checking whether make supports nested variables... yes
checking whether make supports nested variables... (cached) yes
checking whether to enable maintainer-specific portions of Makefiles... no
checking build system type... x86_64-unknown-linux-gnu
checking host system type... x86_64-unknown-linux-gnu
checking shell variables initial values... done
checking if basename works... yes
checking if xargs works... yes
checking for cached host... none
checking for config x86_64-unknown-linux-gnu... no
checking for config x86_64-unknown-linux-gnu... no
checking for config unknown-linux-gnu... no
checking for config unknown-linux-gnu... no
checking for config x86_64-linux-gnu... no
checking for config x86_64-linux-gnu... no
checking for config x86_64-unknown... no
checking for config linux-gnu... found
compiler 'pgcc' is PGI pgcc-17.4-0
compiler 'pgf90' is PGI pgf90-17.4-0
No match to get cxx_version_info for pgc++
checking for config ./config/site-specific/host-Vostro-3800... no
checking for gcc... pgcc
checking whether the C compiler works... no
configure: error: in `/home//hdf5-1.8.17':
configure: error: C compiler cannot create executables
See `config.log' for more details

Any suggestions on how to build will be appreciated.

Best regards,
Ashwin.

2

Hello,
         I already have an existing installation of HDF5 1.8.17 compiled
with gfortran. Now for another project on the same machine I need to
compile with PGI compilers i.e. pgcc , pgc++and pgf90. The installation
directory will be separate as well i.e. /opt/local/lib and
/opt/local/include.

This is how I try to build hdf5

CPPFLAGS=/opt/local/include CXXFLAGS=/opt//local/include
LDFLAGS=/opt//local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure
--prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx.

and when I try doing this I get this error message -

checking for a BSD-compatible install... /usr/bin/install -c
checking whether build environment is sane... yes
[...]

checking for gcc... pgcc

checking whether the C compiler works... no
configure: error: in `/home//hdf5-1.8.17':
configure: error: C compiler cannot create executables
See `config.log' for more details <<<<<<<<<<<<<<<<<<<<

Any suggestions on how to build will be appreciated.

Your flags settings need "-I" or "-L" as appropriate.

If that does't solve your problem, it might help if you provide the OS and
version as well as the compiler version.
Someone on the list may have advice for your particular platform.

It would also be helpful if you could post relevant "more details" from
the config.log file.

···

On 26 June 2017 at 03:42, ashwin .D <winash12@gmail.com> wrote:

--
George N. White III <aa056@chebucto.ns.ca>
Head of St. Margarets Bay, Nova Scotia

3

*On 06/26/2017 01:42 AM, ashwin .D wrote:*

    I already have an existing installation of HDF5 1.8.17 compiled
    with gfortran. Now for another project on the same machine I need
    to compile with PGI compilers i.e. pgcc , pgc++and pgf90. The
    installation directory will be separate as well i.e.
    /opt/local/lib and /opt/local/include.

Ashwin -- here's a script that I use to build HDF5 1.10.1 using the PGI 17.5 compiler.
You might try testing with these components first to see if you can get it to build, and then go backlevel on the pieces that you want to use.

         -- Carl

02.Install (3.54 KB)

···

------------------------------------------------------------------------

This is how I try to build hdf5

    CPPFLAGS=/opt/local/include CXXFLAGS=/opt//local/include
    LDFLAGS=/opt//local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure --prefix=/opt/local --with-zlib=/opt/local
    --enable-fortran --enable-cxx.

and when I try doing this I get this error message -

    checking for a BSD-compatible install... /usr/bin/install -c
    checking whether build environment is sane... yes
    checking for a thread-safe mkdir -p... /bin/mkdir -p
    checking for gawk... no
    checking for mawk... mawk
    checking whether make sets $(MAKE)... yes
    checking whether make supports nested variables... yes
    checking whether make supports nested variables... (cached) yes
    checking whether to enable maintainer-specific portions of
    Makefiles... no
    checking build system type... x86_64-unknown-linux-gnu
    checking host system type... x86_64-unknown-linux-gnu
    checking shell variables initial values... done
    checking if basename works... yes
    checking if xargs works... yes
    checking for cached host... none
    checking for config x86_64-unknown-linux-gnu... no
    checking for config unknown-linux-gnu... no
    checking for config x86_64-linux-gnu... no
    checking for config x86_64-unknown... no
    checking for config linux-gnu... found
    compiler 'pgcc' is PGI pgcc-17.4-0
    compiler 'pgf90' is PGI pgf90-17.4-0
    No match to get cxx_version_info for pgc++
    checking for config ./config/site-specific/host-Vostro-3800... no
    checking for gcc... pgcc
    checking whether the C compiler works... no
    configure: error: in `/home//hdf5-1.8.17':
    configure: error: C compiler cannot create executables
    See `config.log' for more details

Any suggestions on how to build will be appreciated.

-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------

4

George, Matt and Carl.
                                   Thanks much for all the suggestions. I
was able to move past the configure errors by using a combination of all
the tips you recommended. I in fact downloaded the hdf-1.10.1 version and
while running make did not give any problems I have a whole bunch of
redundant macro warnings and which when I looked up is identical to this
one -
https://lists.hdfgroup.org/pipermail/hdf-forum_lists.hdfgroup.org/2010-July/003304.html
.
I have enclosed a trimmed version of my logs. Is it good to proceed forward
or is there something I need to edit with the header files as well ? Note
these warnings do not occur with gfortran.

Best regards,
Ashwin.

logs (65.4 KB)

···

On Tue, Jun 27, 2017 at 3:09 AM, Carl Ponder <cponder@nvidia.com> wrote:

*On 06/26/2017 01:42 AM, ashwin .D wrote:*

I already have an existing installation of HDF5 1.8.17 compiled with
gfortran. Now for another project on the same machine I need to compile
with PGI compilers i.e. pgcc , pgc++and pgf90. The installation directory
will be separate as well i.e. /opt/local/lib and /opt/local/include.

Ashwin -- here's a script that I use to build HDF5 1.10.1 using the PGI
17.5 compiler.
You might try testing with these components first to see if you can get it
to build, and then go backlevel on the pieces that you want to use.

        -- Carl

------------------------------

This is how I try to build hdf5

CPPFLAGS=/opt/local/include CXXFLAGS=/opt//local/include
LDFLAGS=/opt//local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure
--prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx.

and when I try doing this I get this error message -

checking for a BSD-compatible install... /usr/bin/install -c
checking whether build environment is sane... yes
checking for a thread-safe mkdir -p... /bin/mkdir -p
checking for gawk... no
checking for mawk... mawk
checking whether make sets $(MAKE)... yes
checking whether make supports nested variables... yes
checking whether make supports nested variables... (cached) yes
checking whether to enable maintainer-specific portions of Makefiles... no
checking build system type... x86_64-unknown-linux-gnu
checking host system type... x86_64-unknown-linux-gnu
checking shell variables initial values... done
checking if basename works... yes
checking if xargs works... yes
checking for cached host... none
checking for config x86_64-unknown-linux-gnu... no
checking for config x86_64-unknown-linux-gnu... no
checking for config unknown-linux-gnu... no
checking for config unknown-linux-gnu... no
checking for config x86_64-linux-gnu... no
checking for config x86_64-linux-gnu... no
checking for config x86_64-unknown... no
checking for config linux-gnu... found
compiler 'pgcc' is PGI pgcc-17.4-0
compiler 'pgf90' is PGI pgf90-17.4-0
No match to get cxx_version_info for pgc++
checking for config ./config/site-specific/host-Vostro-3800... no
checking for gcc... pgcc
checking whether the C compiler works... no
configure: error: in `/home//hdf5-1.8.17':
configure: error: C compiler cannot create executables
See `config.log' for more details

Any suggestions on how to build will be appreciated.

------------------------------
This email message is for the sole use of the intended recipient(s) and
may contain confidential information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.
------------------------------

5

Sounds promissing.

The post you cite above encounters "PGC-S-..." lines, where the "-S-"
indicates a "Severe" error. Your log has "PGC-I-..." ("Information) and
"PGC-W-..." ("Warning") lines. These generally don't affect the operation
of the compiled code, but are things the developers may want to examine,
especially if they appear when compiling end-user code. You certainly want
to run "make check" (perhaps with HDF5TestExpress=0) and look over the
results carefully. On macOS with PGI I see:

Testing soft denormalized float -> double conversions
*WARNING*
[...]
maximum warnings reached, aborting test...

For hdf5-1.8.17, release_docs/INSTALL has a section "A.2. Building and
testing with PGI compilers" which says:

     set the environment variable CXX to "pgCC -tlocal"

···

On 27 June 2017 at 00:06, ashwin .D <winash12@gmail.com> wrote:

George, Matt and Carl.
                                   Thanks much for all the suggestions. I
was able to move past the configure errors by using a combination of all
the tips you recommended. I in fact downloaded the hdf-1.10.1 version and
while running make did not give any problems I have a whole bunch of
redundant macro warnings and which when I looked up is identical to this
one - https://lists.hdfgroup.org/pipermail/hdf-forum_lists.
hdfgroup.org/2010-July/003304.html.
I have enclosed a trimmed version of my logs. Is it good to proceed
forward or is there something I need to edit with the header files as well
? Note these warnings do not occur with gfortran.

Best regards,
Ashwin.

--
George N. White III <aa056@chebucto.ns.ca>
Head of St. Margarets Bay, Nova Scotia

6

*On 06/26/2017 10:06 PM, ashwin .D wrote:*

    Thanks much for all the suggestions. I was able to move past the
    configure errors by using a combination of all the tips you
    recommended. I in fact downloaded the hdf-1.10.1 version and while
    running make did not give any problems I have a whole bunch of
    redundant macro warnings and which when I looked up is identical
    to this one -
    https://lists.hdfgroup.org/pipermail/hdf-forum_lists.hdfgroup.org/2010-July/003304.html.
    I have enclosed a trimmed version of my logs. Is it good to
    proceed forward or is there something I need to edit with the
    header files as well ? Note these warnings do not occur with
    gfortran.

Ashwin -- I have the same warnings in my logs.
I'm guessing that the same file is being include'd indirectly somehow, in which case the redundant definitions would be identical and not cause problems.
You'd have to ask the HDF5 developers about this, although I'd be interested in knowing the answer myself.
In the link you reference above, I see the warnings but they are in reference to an earlier posting and the reply was not included.

When you ran the regression-tests

    make -i -k check

did they almost all pass?

···

-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------

7

Carl,
           When I ran that command make -i -k check it passed every singe
test. There wasn't a single error whatsoever. By the way I was able to fix
that HDF5 error associated with netCDF as well. There is an environmental
variable (that George hinted at) you have to set and and so I am all set
now with the configuration of HDF5.

Ashwin.

···

On Tue, Jun 27, 2017 at 9:37 PM, Carl Ponder <cponder@nvidia.com> wrote:

*On 06/26/2017 10:06 PM, ashwin .D wrote:*

Thanks much for all the suggestions. I was able to move past the configure
errors by using a combination of all the tips you recommended. I in fact
downloaded the hdf-1.10.1 version and while running make did not give any
problems I have a whole bunch of redundant macro warnings and which when I
looked up is identical to this one - https://lists.hdfgroup.org/
pipermail/hdf-forum_lists.hdfgroup.org/2010-July/003304.html.
I have enclosed a trimmed version of my logs. Is it good to proceed
forward or is there something I need to edit with the header files as well
? Note these warnings do not occur with gfortran.

Ashwin -- I have the same warnings in my logs.
I'm guessing that the same file is being include'd indirectly somehow, in
which case the redundant definitions would be identical and not cause
problems.
You'd have to ask the HDF5 developers about this, although I'd be
interested in knowing the answer myself.
In the link you reference above, I see the warnings but they are in
reference to an earlier posting and the reply was not included.

When you ran the regression-tests

make -i -k check

did they almost all pass?
------------------------------
This email message is for the sole use of the intended recipient(s) and
may contain confidential information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.
------------------------------

8

What was the NetCDF setting? I'll have to check what I'm doing.

         -- Carl

···

On 06/27/2017 11:32 AM, ashwin .D wrote:

    When I ran that command make -i -k check it passed every singe
    test. There wasn't a single error whatsoever. By the way I was
    able to fix that HDF5 error associated with netCDF as well. There
    is an environmental variable (that George hinted at) you have to
    set and and so I am all set now with the configuration of HDF5.

-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------

9

This maybe off topic so there is some background to that issue. After a
successful configuration of HDF5 I went ahead and recompiled NetCDF with
PGI compilers. Following that compilation I went ahead and ran make check
and all of the HDF5 tests failed. So I was not sure whether the HDF5
compilation/configuration was successful or not because of the log messages
that we just discussed. Then it turns out after some searching on the web
you need to have this environment variable set on Linux (and only for
Linux) CPPFLAGS = -DpgiFortran -I/opt/local/include. This then ensures
that all of the HDF5 tests associated with NetCDF successfully complete.

···

On Tue, Jun 27, 2017 at 10:07 PM, Carl Ponder <cponder@nvidia.com> wrote:

On 06/27/2017 11:32 AM, ashwin .D wrote:

When I ran that command make -i -k check it passed every singe test. There
wasn't a single error whatsoever. By the way I was able to fix that HDF5
error associated with netCDF as well. There is an environmental variable
(that George hinted at) you have to set and and so I am all set now with
the configuration of HDF5.

What was the NetCDF setting? I'll have to check what I'm doing.

        -- Carl

------------------------------
This email message is for the sole use of the intended recipient(s) and
may contain confidential information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.
------------------------------

10

n 26 June 2017 at 03:42, ashwin .D <winash12 at gmail.com
<http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org>>
wrote:

* Hello,

*>* I already have an existing installation of HDF5 1.8.17 compiled
*>* with gfortran. Now for another project on the same machine I need to
*>* compile with PGI compilers i.e. pgcc , pgc++and pgf90. The installation
*>* directory will be separate as well i.e. /opt/local/lib and
*>* /opt/local/include.
*>>* This is how I try to build hdf5
*>>* CPPFLAGS=/opt/local/include CXXFLAGS=/opt//local/include
*>* LDFLAGS=/opt//local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure
*>* --prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx.
*>>* and when I try doing this I get this error message -
*>>* checking for a BSD-compatible install... /usr/bin/install -c
*>* checking whether build environment is sane... yes
*>* [...]
*>checking for gcc... pgcc

* checking whether the C compiler works... no

*>* configure: error: in `/home//hdf5-1.8.17':
*>* configure: error: C compiler cannot create executables
*>* See `config.log' for more details <<<<<<<<<<<<<<<<<<<<
*>>* Any suggestions on how to build will be appreciated.
*>>>Your flags settings need "-I" or "-L" as appropriate.

If that does't solve your problem, it might help if you provide the OS and
version as well as the compiler version.
Someone on the list may have advice for your particular platform.

It would also be helpful if you could post relevant "more details" from
the config.log file.

Sorry for asking the question. In this command line for configure

how would I add -I and -L

*CPPFLAGS=/opt/local/include CXXFLAGS=/opt/local/include
** LDFLAGS=/opt/local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure
** --prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx.

config.log (25.3 KB)

···

*

The compiler version is

pgcc 17.4-0 64-bit target on x86-64 Linux -tp haswell
PGI Compilers and Tools
Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.

OS version as given by uname Ubuntu 16.04 LTS

Linux a-Vostro-3800 4.4.0-81-generic #104-Ubuntu SMP Wed Jun 14 08:17:06
UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

11

You need to do something like:

   CPPFLAGS='-I/opt/local/include' CXXFLAGS='-I/opt/local/include' LDFLAGS='-L/opt/local/lib'

For include directories, use -I, for library locations use -L. The flags are literal and won't add them automatically like autotools might through a --with-xxx

Matt

···

On 06/26/2017 12:52 PM, ashwin .D wrote:

n 26 June 2017 at 03:42, ashwin .D <winash12 at gmail.com <http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org>> wrote:

/Hello, />/I already have an existing installation of HDF5 1.8.17 compiled />/with gfortran. Now for another project on the same machine I need to />/compile with PGI compilers i.e. pgcc , pgc++and pgf90. The installation />/directory will be separate as well i.e. /opt/local/lib and />//opt/local/include. />//>/This is how I try to build hdf5 />//>/CPPFLAGS=/opt/local/include CXXFLAGS=/opt//local/include />/LDFLAGS=/opt//local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure />/--prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx. />//>/and when I try doing this I get this error message - />//>/checking for a BSD-compatible install... /usr/bin/install -c />/checking whether build environment is sane... yes />/[...] />//checking for gcc... pgcc
/checking whether the C compiler works... no />/configure: error: in `/home//hdf5-1.8.17': />/configure: error: C compiler cannot create executables />/See `config.log' for more details <<<<<<<<<<<<<<<<<<<< />//>/Any suggestions on how to build will be appreciated. />//>//>Your flags settings need "-I" or "-L" as appropriate.

If that does't solve your problem, it might help if you provide the OS and
version as well as the compiler version.
Someone on the list may have advice for your particular platform.

It would also be helpful if you could post relevant "more details" from
the config.log file.

Sorry for asking the question. In this command line for configure

how would I add -I and -L

/CPPFLAGS=/opt/local/include CXXFLAGS=/opt/local/include //LDFLAGS=/opt/local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure //--prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx. /

--
Matt Thompson, SSAI, Sr Scientific Programmer/Analyst
NASA GSFC, Global Modeling and Assimilation Office
Code 610.1, 8800 Greenbelt Rd, Greenbelt, MD 20771
Phone: 301-614-6712 Fax: 301-614-6246
http://science.gsfc.nasa.gov/sed/bio/matthew.thompson

12

n 26 June 2017 at 03:42, ashwin .D <winash12 at gmail.com <http://lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org>> wrote:

>* Hello,
*>* I already have an existing installation of HDF5 1.8.17 compiled
*>* with gfortran. Now for another project on the same machine I need to
*>* compile with PGI compilers i.e. pgcc , pgc++and pgf90. The installation
*>* directory will be separate as well i.e. /opt/local/lib and
*>* /opt/local/include.
*>>* This is how I try to build hdf5
*>>* CPPFLAGS=/opt/local/include CXXFLAGS=/opt//local/include
*>* LDFLAGS=/opt//local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure
*>* --prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx.
*

Seeing (below) that you are using linux, you probably don't want to set
CPPFLAGS, CXXFLAGS, or
LDFLAGS at all. You should not need *--with-zlib=/opt/local *either, but
you will need to
install the libz development package. An exception would be if your site
policy doesn't allow you
to install developer versions of system libriaries but did allow you to
install the libraries in /opt/local/.

>* and when I try doing this I get this error message -
*>>* checking for a BSD-compatible install... /usr/bin/install -c
*>* checking whether build environment is sane... yes
*>* [...]
*>checking for gcc... pgcc
>* checking whether the C compiler works... no
*>* configure: error: in `/home//hdf5-1.8.17':
*>* configure: error: C compiler cannot create executables
*>* See `config.log' for more details <<<<<<<<<<<<<<<<<<<<
*>>* Any suggestions on how to build will be appreciated.
*>>>Your flags settings need "-I" or "-L" as appropriate.

>If that does't solve your problem, it might help if you provide the OS and
>version as well as the compiler version.
>Someone on the list may have advice for your particular platform.

>It would also be helpful if you could post relevant "more details" from
>the config.log file.

Sorry for asking the question. In this command line for configure

how would I add -I and -L

*CPPFLAGS=/opt/local/include CXXFLAGS=/opt/local/include
** LDFLAGS=/opt/local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure
** --prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx.
*

Like Matt said, but be sure these are the correct settings for your system.

The compiler version is

pgcc 17.4-0 64-bit target on x86-64 Linux -tp haswell
PGI Compilers and Tools
Copyright (c) 2017, NVIDIA CORPORATION. All rights reserved.

OS version as given by uname Ubuntu 16.04 LTS

Linux a-Vostro-3800 4.4.0-81-generic #104-Ubuntu SMP Wed Jun 14 08:17:06
UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

Your config.log file has errors like "/usr/bin/ld: cannot find
/opt/local/include: File format not recognized". This is what
I would expect to see with missing "-I".

···

On 26 June 2017 at 13:52, ashwin .D <winash12@gmail.com> wrote:

--
George N. White III <aa056@chebucto.ns.ca>
Head of St. Margarets Bay, Nova Scotia

13

* George, Matt and Carl.

*>* Thanks much for all the suggestions. I
*>* was able to move past the configure errors by using a combination of all
*>* the tips you recommended. I in fact downloaded the hdf-1.10.1 version and
*>* while running make did not give any problems I have a whole bunch of
*>* redundant macro warnings and which when I looked up is identical to this
*>* one - https://lists.hdfgroup.org/pipermail/hdf-forum_lists.
<https://lists.hdfgroup.org/pipermail/hdf-forum_lists.>
*>* hdfgroup.org/2010-July/003304.html
<http://hdfgroup.org/2010-July/003304.html>.
*>* I have enclosed a trimmed version of my logs. Is it good to proceed
*>* forward or is there something I need to edit with the header files as well
*>* ? Note these warnings do not occur with gfortran.
*>>* Best regards,
*>* Ashwin.
*>

Sounds promissing.

The post you cite above encounters "PGC-S-..." lines, where the "-S-"
indicates a "Severe" error. Your log has "PGC-I-..." ("Information) and
"PGC-W-..." ("Warning") lines. These generally don't affect the operation
of the compiled code, but are things the developers may want to examine,
especially if they appear when compiling end-user code. You certainly want
to run "make check" (perhaps with HDF5TestExpress=0) and look over the
results carefully. On macOS with PGI I see:

Testing soft denormalized float -> double conversions
*WARNING*
[...]
maximum warnings reached, aborting test...

For hdf5-1.8.17, release_docs/INSTALL has a section "A.2. Building and
testing with PGI compilers" which says:

> set the environment variable CXX to "pgCC -tlocal"

Thanks I will definitely try that. But I have to let Carl and you know that
I do not have a pgCC on my box. I only have a pgc++. I checked for a pgCC
under my installation and I could not find it.
What about modifying the header files as indicated in that post I linked ?
Are they not required ?

···

On Tue, Jun 27, 2017 at 8:36 AM, ashwin .D <winash12@gmail.com> wrote:

George, Matt and Carl.
                                   Thanks much for all the suggestions. I
was able to move past the configure errors by using a combination of all
the tips you recommended. I in fact downloaded the hdf-1.10.1 version and
while running make did not give any problems I have a whole bunch of
redundant macro warnings and which when I looked up is identical to this
one - https://lists.hdfgroup.org/pipermail/hdf-forum_lists.
hdfgroup.org/2010-July/003304.html.
I have enclosed a trimmed version of my logs. Is it good to proceed
forward or is there something I need to edit with the header files as well
? Note these warnings do not occur with gfortran.

Best regards,
Ashwin.

On Tue, Jun 27, 2017 at 3:09 AM, Carl Ponder <cponder@nvidia.com> wrote:

*On 06/26/2017 01:42 AM, ashwin .D wrote:*

I already have an existing installation of HDF5 1.8.17 compiled with
gfortran. Now for another project on the same machine I need to compile
with PGI compilers i.e. pgcc , pgc++and pgf90. The installation directory
will be separate as well i.e. /opt/local/lib and /opt/local/include.

Ashwin -- here's a script that I use to build HDF5 1.10.1 using the PGI
17.5 compiler.
You might try testing with these components first to see if you can get
it to build, and then go backlevel on the pieces that you want to use.

        -- Carl

------------------------------

This is how I try to build hdf5

CPPFLAGS=/opt/local/include CXXFLAGS=/opt//local/include
LDFLAGS=/opt//local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure
--prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx.

and when I try doing this I get this error message -

checking for a BSD-compatible install... /usr/bin/install -c
checking whether build environment is sane... yes
checking for a thread-safe mkdir -p... /bin/mkdir -p
checking for gawk... no
checking for mawk... mawk
checking whether make sets $(MAKE)... yes
checking whether make supports nested variables... yes
checking whether make supports nested variables... (cached) yes
checking whether to enable maintainer-specific portions of Makefiles... no
checking build system type... x86_64-unknown-linux-gnu
checking host system type... x86_64-unknown-linux-gnu
checking shell variables initial values... done
checking if basename works... yes
checking if xargs works... yes
checking for cached host... none
checking for config x86_64-unknown-linux-gnu... no
checking for config x86_64-unknown-linux-gnu... no
checking for config unknown-linux-gnu... no
checking for config unknown-linux-gnu... no
checking for config x86_64-linux-gnu... no
checking for config x86_64-linux-gnu... no
checking for config x86_64-unknown... no
checking for config linux-gnu... found
compiler 'pgcc' is PGI pgcc-17.4-0
compiler 'pgf90' is PGI pgf90-17.4-0
No match to get cxx_version_info for pgc++
checking for config ./config/site-specific/host-Vostro-3800... no
checking for gcc... pgcc
checking whether the C compiler works... no
configure: error: in `/home//hdf5-1.8.17':
configure: error: C compiler cannot create executables
See `config.log' for more details

Any suggestions on how to build will be appreciated.

------------------------------
This email message is for the sole use of the intended recipient(s) and
may contain confidential information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.
------------------------------

14

I encountered no problems with make check in HDF5. So I went ahead and
installed it under /opt/local. Then I went downstream and compiled
netcdf-4.4.1 while pointing to the HDF5 installation under /opt/local and
could compile that as well. But when I did a make check on that one I got a
lot of HDF 5 related errors. So obviously my HDF5 compilation is not a done
deal yet.

···

On Tue, Jun 27, 2017 at 4:52 PM, ashwin .D <winash12@gmail.com> wrote:

>* George, Matt and Carl.
*>* Thanks much for all the suggestions. I
*>* was able to move past the configure errors by using a combination of all
*>* the tips you recommended. I in fact downloaded the hdf-1.10.1 version and
*>* while running make did not give any problems I have a whole bunch of
*>* redundant macro warnings and which when I looked up is identical to this
*>* one - https://lists.hdfgroup.org/pipermail/hdf-forum_lists. <https://lists.hdfgroup.org/pipermail/hdf-forum_lists.>
*>* hdfgroup.org/2010-July/003304.html <http://hdfgroup.org/2010-July/003304.html>.
*>* I have enclosed a trimmed version of my logs. Is it good to proceed
*>* forward or is there something I need to edit with the header files as well
*>* ? Note these warnings do not occur with gfortran.
*>>* Best regards,
*>* Ashwin.
*>
>Sounds promissing.

>The post you cite above encounters "PGC-S-..." lines, where the "-S-"
>indicates a "Severe" error. Your log has "PGC-I-..." ("Information) and
>"PGC-W-..." ("Warning") lines. These generally don't affect the operation
>of the compiled code, but are things the developers may want to examine,
>especially if they appear when compiling end-user code. You certainly want
>to run "make check" (perhaps with HDF5TestExpress=0) and look over the
>results carefully. On macOS with PGI I see:

>Testing soft denormalized float -> double conversions
>*WARNING*
>[...]
>maximum warnings reached, aborting test...

>For hdf5-1.8.17, release_docs/INSTALL has a section "A.2. Building and
>testing with PGI compilers" which says:

> set the environment variable CXX to "pgCC -tlocal"

Thanks I will definitely try that. But I have to let Carl and you know
that I do not have a pgCC on my box. I only have a pgc++. I checked for a
pgCC under my installation and I could not find it.
What about modifying the header files as indicated in that post I linked ?
Are they not required ?

On Tue, Jun 27, 2017 at 8:36 AM, ashwin .D <winash12@gmail.com> wrote:

George, Matt and Carl.
                                   Thanks much for all the suggestions.
I was able to move past the configure errors by using a combination of all
the tips you recommended. I in fact downloaded the hdf-1.10.1 version and
while running make did not give any problems I have a whole bunch of
redundant macro warnings and which when I looked up is identical to this
one - https://lists.hdfgroup.org/pipermail/hdf-forum_lists.hdfgrou
p.org/2010-July/003304.html.
I have enclosed a trimmed version of my logs. Is it good to proceed
forward or is there something I need to edit with the header files as well
? Note these warnings do not occur with gfortran.

Best regards,
Ashwin.

On Tue, Jun 27, 2017 at 3:09 AM, Carl Ponder <cponder@nvidia.com> wrote:

*On 06/26/2017 01:42 AM, ashwin .D wrote:*

I already have an existing installation of HDF5 1.8.17 compiled with
gfortran. Now for another project on the same machine I need to compile
with PGI compilers i.e. pgcc , pgc++and pgf90. The installation directory
will be separate as well i.e. /opt/local/lib and /opt/local/include.

Ashwin -- here's a script that I use to build HDF5 1.10.1 using the PGI
17.5 compiler.
You might try testing with these components first to see if you can get
it to build, and then go backlevel on the pieces that you want to use.

        -- Carl

------------------------------

This is how I try to build hdf5

CPPFLAGS=/opt/local/include CXXFLAGS=/opt//local/include
LDFLAGS=/opt//local/lib CC=pgcc CXX=pgc++ FC=pgf90 F77=pgf77 ./configure
--prefix=/opt/local --with-zlib=/opt/local --enable-fortran --enable-cxx.

and when I try doing this I get this error message -

checking for a BSD-compatible install... /usr/bin/install -c
checking whether build environment is sane... yes
checking for a thread-safe mkdir -p... /bin/mkdir -p
checking for gawk... no
checking for mawk... mawk
checking whether make sets $(MAKE)... yes
checking whether make supports nested variables... yes
checking whether make supports nested variables... (cached) yes
checking whether to enable maintainer-specific portions of Makefiles...
no
checking build system type... x86_64-unknown-linux-gnu
checking host system type... x86_64-unknown-linux-gnu
checking shell variables initial values... done
checking if basename works... yes
checking if xargs works... yes
checking for cached host... none
checking for config x86_64-unknown-linux-gnu... no
checking for config x86_64-unknown-linux-gnu... no
checking for config unknown-linux-gnu... no
checking for config unknown-linux-gnu... no
checking for config x86_64-linux-gnu... no
checking for config x86_64-linux-gnu... no
checking for config x86_64-unknown... no
checking for config linux-gnu... found
compiler 'pgcc' is PGI pgcc-17.4-0
compiler 'pgf90' is PGI pgf90-17.4-0
No match to get cxx_version_info for pgc++
checking for config ./config/site-specific/host-Vostro-3800... no
checking for gcc... pgcc
checking whether the C compiler works... no
configure: error: in `/home//hdf5-1.8.17':
configure: error: C compiler cannot create executables
See `config.log' for more details

Any suggestions on how to build will be appreciated.

------------------------------
This email message is for the sole use of the intended recipient(s) and
may contain confidential information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.
------------------------------

15

I encountered no problems with make check in HDF5. So I went ahead and
installed it under /opt/local. Then I went downstream and compiled
netcdf-4.4.1 while pointing to the HDF5 installation under /opt/local and
could compile that as well. But when I did a make check on that one I got a
lot of HDF 5 related errors. So obviously my HDF5 compilation is not a done
deal yet.

>* George, Matt and Carl.
*>* Thanks much for all the suggestions. I
*>* was able to move past the configure errors by using a combination of all
*>* the tips you recommended. I in fact downloaded the hdf-1.10.1 version and
*>* while running make did not give any problems I have a whole bunch of
*>* redundant macro warnings and which when I looked up is identical to this
*>* one - https://lists.hdfgroup.org/pipermail/hdf-forum_lists. <https://lists.hdfgroup.org/pipermail/hdf-forum_lists.>
*>* hdfgroup.org/2010-July/003304.html <http://hdfgroup.org/2010-July/003304.html>.
*>* I have enclosed a trimmed version of my logs. Is it good to proceed
*>* forward or is there something I need to edit with the header files as well
*>* ? Note these warnings do not occur with gfortran.
*>>* Best regards,
*>* Ashwin.
*>
>Sounds promissing.

>The post you cite above encounters "PGC-S-..." lines, where the "-S-"
>indicates a "Severe" error. Your log has "PGC-I-..." ("Information) and
>"PGC-W-..." ("Warning") lines. These generally don't affect the operation
>of the compiled code, but are things the developers may want to examine,
>especially if they appear when compiling end-user code. You certainly want
>to run "make check" (perhaps with HDF5TestExpress=0) and look over the
>results carefully. On macOS with PGI I see:

>Testing soft denormalized float -> double conversions
>*WARNING*
>[...]
>maximum warnings reached, aborting test...

>For hdf5-1.8.17, release_docs/INSTALL has a section "A.2. Building and
>testing with PGI compilers" which says:

> set the environment variable CXX to "pgCC -tlocal"

Thanks I will definitely try that. But I have to let Carl and you know
that I do not have a pgCC on my box. I only have a pgc++. I checked for a
pgCC under my installation and I could not find it.
What about modifying the header files as indicated in that post I linked
?

The changes to the header files should just hide the redundant defines, so
shouldn't affect the compilation.

It looks like the "pgCC -tlocal" usage is for an older version of the PGI
compilers. From an old PGI manual

"On Linux and macOS, the GNU-compatible pgc++ compiler uses the GNU g++
header files and Standard

Template Library (STL) directly. The versions used are dependent on the
version of the GNU compilers

installed on your system, or specified when makelocalrc was run during
installation of the PGI compilers. "

Are they not required ?

Hard to answer without seeing the errors for the netcdf compile, but in
general, once you have hdf5 libraries
that pass the tests, problems using the libraries are not fixed by changes
in the hdf5 build. For those you may
want to post on a netcdf forum. Although many people reading this list use
netcdf, others who run into the
same problems you are having when building netcdf will be looking for
answers in netcdf forums, so for the
sake of future users it is best to keep hdf5 separate from netcdf.

···

On 27 June 2017 at 08:24, ashwin .D <winash12@gmail.com> wrote:

On Tue, Jun 27, 2017 at 4:52 PM, ashwin .D <winash12@gmail.com> wrote:

on the web, the -tlocal option is to make separate functions of each template instantiaion. The current pgi17ug_x64.pdf (page 135) says: