Red Hat Bugzilla – Bug 171268
Review Request: kdissert
Last modified: 2007-11-30 17:11:15 EST
Spec Name or Url: http://linux.duke.edu/~icon/misc/fe/kdissert.spec
SRPM Name or Url: http://linux.duke.edu/~icon/misc/fe/kdissert-1.0.5-0.1.fc4.src.rpm
kdissert is a mindmapping-like tool to help students to produce complicated
documents very quickly and efficiently : presentations, dissertations,
thesis, reports, etc. The concept is innovative : mindmaps produced using
kdissert are processed to output near-ready-to-use documents. While
targetted mostly at students, kdissert can also help teachers, decision
makers, engineers and businessmen.
Any special reason for installing the .desktop file into the kde sub-directory
of .../applications? I have not seen this in any other FE package. I don't
really care just wanted to know if there is a special reason for this?
Created attachment 120184 [details]
The icon does not have to be copied to the pixmap directory if it is only for
the menu entry. To see the icon I have added the gtk-update-icon-cache to the
scripts and as the desktop file has a mime type entry the desktop database is
also updated in the scriptlets.
I just saw that the patch has to be applied with -R. Sorry.
The .desktop file is where the upstream installs it. I've checked with the stuff
in core, and there's quite a few KDE-related .desktop files in the kde subdir or
applications, so it seems it's a common thing. It doesn't interfere with the way
the application works, so I have left the way it is.
Thanks for the patch, it's been applied. See:
rpmlint complains about rpath. Can be removed with
sed -i -e "/env.KDEuse(\"environ rpath\")/d" SConstruct
before calling scons
Same place: http://linux.duke.edu/~icon/misc/fe/
md5sum of the tarball in the SRPM and the upstream tarball do not match.
According to diff the content seems to be same.
md5sum from SRPM:
Could you make a new version of the SRPM with the new tarball?
Hmm... This is odd. I recall distinctly checking the md5sum of the tarball
before packaging it. It's on my packaging checklist. I'll email the maintainers
to see what's up.
From the author:
Checksum should be 18ff5d04d633cf3b4e3fbf869c18dd2f
There was a mistake in the archive (documentation
installation) so the tarball was updated some time
I've updated the SRPM to contain the matching tarball. See:
- MUST: Packages must NOT contain any .la libtool archives, these should be
removed in the spec.
$ rpm -ql kdissert | grep la$
Seeing the .la discussions all the time you should probably remove these files.
It doesn't work without them. Try it. When .la files are removed, exporting
Yes, after further investigation, making it not require .la files for exporting
plugins would require patching and be generally counter-productive, since it
would diverge from upstream. I think the opposition to .la files is only when
they are included as part of -devel packages (and thus never used), not when
they are required for program functioning during runtime.
* rpmlint is happy
* builds in mock
* clean installation and removal
* spec looks good
* scripts are sane
* works as expected
* source matches upstream
* the .la files are required during runtime and therefore no blocker
> It doesn't work without them. Try it. When .la files are removed, exporting
> tools break.
Is this true of FC5 too?
I'm pretty sure. I've tried playing around to patch the requirement for .la
files out, but opening .so files didn't work, potentially because of a bug in
KDE (i.e. when you tell it to dlopen "libfoo", it will look for libfoo.la, and
ignore libfoo.so. If you specifically instruct to open "libfoo.so", it will fail
because it will tell that "init_libfoo.so not found", though it should have been
trying to find "init_libfoo" in the .so file as well. It's possible that I'm
just dumb. I'm not a C++/KDE programmer).