In the current FC devel pango, the line : BuildRequires: cairo >= %{cairo_version} Needs to be changed to : BuildRequires: cairo-devel >= %{cairo_version} Otherwise the package might get built without cairo support, and gtk2 will then refuse to rebuild. (note that the requirement in the pango-devel sub-package is correct)
Note also that in the spec file, this : Prereq: libXrender, libX11, libXext, libXft Is not needed, as those library dependencies will get picked up automatically by rpm. Since there is no specific version requirement (unlike freetype, gtk2 and cairo here). One more note, these lines are redundant and the first could be removed : Prereq: freetype >= %{freetype_version} Requires: freetype >= %{freetype_version} (since no %pre scriplet requires stuff contained in freetype) Maybe you'd be interested in a global spec file patch to clean it all up?
Sure, a patch would be great. I think the versioned freetype req may be there to avoid some bugs in older versions, though
While looking some more at the spec file, I started wondering about if there could be a way to simplify the whole $host and -32/-64 binary install process. The part that bothers me most is that you end up with a pretty much useless shell script in %post, since it tests a hardcoded string... It would be much cleaner to get the "final" $host as an rpm macro in order to just use it everywhere in the spec (including %post and %files). Another quick suggestion : %if %{_lib} == lib64 /usr/bin/pango-querymodules-64 %else /usr/bin/pango-querymodules-32 %endif This way, once again, there is no test against the (now) hardcoded $host made in %post, but instead the correct script is directly called. Comments? Should I go ahead and try and implement those changes?
I fixed the cairo buildrequires
OK, I guess you're not too keen taking the risk to break the whole 32/64bit script thing, which I perfectly understand :-)