[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: Request for Review: hdf5 (was Re: hfd5 license)

Tom 'spot' Callaway wrote:

Did you send this through the https://admin.fedora.redhat.com/accounts/ system?

I had forgotten to request the agreement. That has been sent in now so I should be getting closer...

I'll check out the package tomorrow. Looks very close to what I have. Main difference is that I build hdf5, libhdf5, and libhdf5-devel packages. I like this format as it separates the binaries and libraries, but don't know if there is fedora policy on this.

IMHO, there is no reason to separate the binaries into their own
package. In order to use them, you'll need to have the libraries
installed. Precedence in both Fedora Core and Extras has the binaries
going in the base package, and static libs, headers, and .so files in
the -devel package.

In hdf5, the sum of the binaries comes out to roughly 544K. Perhaps if
these binaries were in the >10M category, they would merit their own
package to save on download space...

My only thought on this is that someday it might make the availability of multiple versions of the libraries easier, but that appears unlikely.


- added %check section with "make check".  It fails here:

Testing h5diff file5.h5 file6.h5 -v *FAILED*
Expected result (*.txt) differs from actual result (*.out)
*** ./../testfiles/h5diff_70.txt 2004-07-21 11:50:11.000000000 -0600
--- ../testfiles/h5diff_70.out 2005-06-29 14:49:17.000000000 -0600
*** 10,58 ****
Dataset: </dset> and </dset>
</dset> and </dset> are empty datasets
Group: </g1> and </g1>
! Attribute: <string of </g1>> and <string of </g1>>
! position string of </g1> string of </g1> difference
[ 0 ] a z
[ 0 ] b z
[ 1 ] d z
[ 1 ] e z
4 differences found

Basically, the output of h5diff is different than expected. Instead of printing "<string of </g1>>" it prints "< of </>>". It seems some string buffer must get overritten. Could not reproduce when compiled with -g.

- changed %install to use %makeinstall
- don't we want some %docs?
- we might want to specify "--enable-threadsafe --with-pthread" to configure. No realy opinion on this, but I could see it helping with openmp programs. Note that is has no effect on make check above.

Here's a diff:

--- hdf5-tom.spec 2005-06-29 11:48:44.000000000 -0600
+++ hdf5.spec 2005-06-29 15:21:50.000000000 -0600
@@ -38,16 +38,17 @@
# the Fortran parts are hitting GCC bug 17917, add --enable-fortran when it gets fixed.
-%configure --with-ssl --enable-cxx
+%configure --with-ssl --enable-cxx --enable-threadsafe --with-pthread

-mkdir -p $RPM_BUILD_ROOT/%{_docdir}
-mv $RPM_BUILD_ROOT/usr/doc/* $RPM_BUILD_ROOT/%{_docdir}
+%makeinstall docdir=${RPM_BUILD_ROOT}%{_docdir}
rm -rf $RPM_BUILD_ROOT/%{_libdir}/*.la $RPM_BUILD_ROOT/%{_libdir}/*.settings

+#make check

@@ -57,6 +58,8 @@

+%doc COPYING MANIFEST README.txt release_docs/RELEASE.txt
+%doc release_docs/HISTORY.txt doc/html
@@ -82,7 +85,12 @@

-* Mon Jun 27 2005 Tom "spot" Callaway <tcallawa redhat com> 1.6.4-1
+* Wed Jun 29 2005 Orion Poplawski <orion cora nwra com> 1.6.4-3
+- Add --enable-threads --with-pthreads to configure
+- Add %check (commented out 'till it works)
+- Use %makeinstall
+* Mon Jun 27 2005 Tom "spot" Callaway <tcallawa redhat com> 1.6.4-2
 - remove szip from spec, since szip license doesn't meet Fedora standards

 * Sun Apr 3 2005 Tom "spot" Callaway <tcallawa redhat com> 1.6.4-1

- Orion

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]