A few things that I have in my local version of the ebuild: * use of python eclass * parallel builds fine here * USE_EINSTALL doesn't seem necessary * updated dependencies * still needs the linguas patch * some doc and mime stuff that I kept but the ebuild in portage seems to have dropped (notably the still working man page). posting the diff in a minute.
Created attachment 147788 [details, diff] gdesklets-core-0.36.ebuild.patch here it goes.
(In reply to comment #0) > * use of python eclass Not entirely sure why you go through the trouble. Can you explain? > * parallel builds fine here It doesn't on all systems. See bug 215300 and search the forums, for example. My theory is that you don't have a processor with multiple cores. > * USE_EINSTALL doesn't seem necessary Technically it doesn't need to be set, but in the gnome2 eclass it checks this variable to decide on how to build the package. I chose to include it to explicitly call out this decision. > * updated dependencies Until bugs are filed that require the dependencies to be changed, I won't change them. According to upstream, the dependencies haven't changed. For example, it doesn't directly depend on gtk+ and some of the ones you removed were added as bugfixes. > * still needs the linguas patch This is apparently true. Fixed. > * some doc and mime stuff that I kept but the ebuild in portage seems to have > dropped (notably the still working man page). Manpage is automatically installed now by make, develbook has not been regenerated for this version. The entire mime directory had to be removed, see bug 211290, bug 151206.
(In reply to comment #2) > (In reply to comment #0) > > * use of python eclass > > Not entirely sure why you go through the trouble. Can you explain? Generating .py[co] at pkg_postinst time is better because it means that these "optimized" python files are generated on the target machine (think binpkg only boxes which might have not exactly the same python version). > > * parallel builds fine here > > It doesn't on all systems. See bug 215300 and search the forums, for example. > My theory is that you don't have a processor with multiple cores. I have a core2 duo, amd64 thurion x2, pentium D, and distcc farm :) Anyway the mentionned bug is happening in install-local-data and to my knowledge it was fixed in 0.36. Now I can double check that when I get at work on monday. > > * USE_EINSTALL doesn't seem necessary > > Technically it doesn't need to be set, but in the gnome2 eclass it checks this > variable to decide on how to build the package. I chose to include it to > explicitly call out this decision. > > > * updated dependencies > > Until bugs are filed that require the dependencies to be changed, I won't > change them. According to upstream, the dependencies haven't changed. For > example, it doesn't directly depend on gtk+ and some of the ones you removed > were added as bugfixes. I've updated the deps by reading the configure.in, so... ack for removed deps. > > > * some doc and mime stuff that I kept but the ebuild in portage seems to have > > dropped (notably the still working man page). > > Manpage is automatically installed now by make, develbook has not been > regenerated for this version. The entire mime directory had to be removed, see > bug 211290, bug 151206. > ack Thanks for reviewing this.
Thanks for your help. I've incorporated your python compilation changes into 0.36-r1.
thanks again. Oh and about the parallel build issue, I've finally stomped on it. It happened to me in make install, maybe I'll get time to fix this and report :)
