Re: [Plplot-devel] PLplot Fortran bindings

[Alan W. I. <ir...@be...>]

Hi Arjen:

On 2016-04-28 08:55-0000 Arjen Markus wrote:

> I made the small change I suggested and the results for Cygwin and
MinGW64 look fine - see the attached files.

Thanks for those two test report tarballs.  Here is my analysis of
those results.

1. Cygwin

After unpacking your report tarball, I didn't spot anything unusual in
your comprehensive test script arguments except for

--build_command "make -j4"
--ctest_command "ctest -j4"

I believe you have sometimes found those parallel options to be
problematic for Cygwin in the past so I would drop them from
now on (even though they caused no issues for you for this particular
comprehensive test).

Furthermore,
comprehensive_test.sh.out showed no errors;

grep -i error */*/output_tree/*.out

and

grep -B1 -A3 "Missing examples" */*/output_tree/*.out |less

gave perfect results; and there were no significant warnings in 
shared/noninteractive/output_tree/make_noninteractive.out or
shared/noninteractive/output_tree/cmake.out.

However, a final check

grep local shared/noninteractive/output_tree/cmake.out

yielded

-- FindShapelib: Found shapelib header directory, /usr/local/include, and library, /usr/lib/libshp.a.

indicating a potentially serious inconsistency between the Cygwin
location of /usr/lib/libshp.a and the headers of some local non-Cygwin
version of shapelib.  That inconsistency obviously caused no trouble
for your current tests, but it is an "accident waiting to happen"
which might occur any time you upgrade the Cygwin version of shapelib.

A Cygwin package search reveals there is no Cygwin package (at least
for up-to-date Cygwin) that includes libshp.a.  So I wonder if
/usr/lib/libshp.a above is from some older version of Cygwin?

Also, the relevant header file being searched for is called
shapefil.h, and that (and /usr/lib/libshp.dll.a which is the relevant
Cygwin shapefile library that you should have found above) should be
installed as part of the libshp-devel-1.3.0-1 package.

Anyhow, I suspect this issue will go away once you install the
libshp-devel-1.3.0-1 Cygwin package (which should automatically
get rid of the above /usr/lib/libshp.a if that is from some
previous version of Cygwin).

This may or may not be relevant to the above issue, but I notice in
comprehensive_env.out that your PATH variable includes /usr/local/bin
which significantly affects CMake find capabilities, i.e., in general,
if a PATH component is PREFIX/bin, CMake will search other standard
locations, (e.g., PREFIX/include) associated with that PREFIX for
include files (and similarly for libraries).  Anyhow, you only want to
depend on pure Cygwin software for these tests, so I suggest you
remove /usr/local/bin from your PATH for all subsequent Cygwin tests.

In sum, your Fortran test report is encouraging for this platform, but
I will wait for your next Cygwin test (of my latest Fortran commit)
with the parallel make and ctest options dropped, and this shapelib
and /usr/local/bin issue fixed before posting your Cygwin
Fortran-constrained comprehensive test reports to our Wiki.

2. MinGW-w64/MSYS2

After unpacking your report tarball, I didn't spot anything unusual in
your comprehensive test script arguments.  For example, there is no
parallel build or ctest option which is correct for this platform (and
also Cygwin, see above comment).

Furthermore,
comprehensive_test.sh.out showed no errors; and

grep -i error */*/output_tree/*.out

and

grep -B1 -A3 "Missing examples" */*/output_tree/*.out |less

gave perfect results.

However, shared/noninteractive/output_tree/cmake.out showed that
libqhull was missing which adds some important core capabilities to
PLplot. If you look at the (dated) package list at
<https://sourceforge.net/p/msys2/wiki/Packages/> MinGW-w64/MSYS2 does
include a qhull-related package.  So you should install the latest
version of that package to address this issue.  Also, I highly
recommend (if you haven't done so already) you reinstall
MinGW-w64/MSYS2 from scratch following the procedure given in
<https://sourceforge.net/p/msys2/wiki/MSYS2 installation/> to get
access to the latest much-simplified update procedure before
installing PLplot dependencies including the qhull-related package.

Furthermore,

grep local shared/noninteractive/output_tree/cmake.out

found no "local" issue, and no significant warnings appeared
in shared/noninteractive/output_tree/make_noninteractive.out other
than the expected warnings about libqhull-related functionality
being missing.

In sum, your Fortran test report is encouraging on this platform, but
I will wait for your next MinGW-w64/MSYS2 test (of my latest Fortran
commit) with reinstalled MinGW-w64/MSYS2 (if you haven't done that
already) and the qhull-related package installed, before posting your
MinGW-w64/MSYS2 Fortran-constrained comprehensive test reports to our
Wiki.

Alan
__________________________
Alan W. Irwin

Astronomical research affiliation with Department of Physics and Astronomy,
University of Victoria (astrowww.phys.uvic.ca).

Programming affiliations with the FreeEOS equation-of-state
implementation for stellar interiors (freeeos.sf.net); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); the libLASi project
(unifont.org/lasi); the Loads of Linux Links project (loll.sf.net);
and the Linux Brochure Project (lbproject.sf.net).
__________________________

Linux-powered Science
__________________________