Hi arches, with the introduction of eog-2.24 in tree, you just lost keywords on it because of the new exempi dependency. Please either use.mask the xmp USE flag or keyword media-libs/exempi and readd your keyword to eog-2.24 ebuild. It is expected that some other ebuilds will require it as well (nautilus, tracker) for which we'll fill separate bugs in due time. Thanks in advance.
Hi arches, it's me again, following previous comment, you just loosed keyword on nautilus 2.24. gnome-base/nautilus/nautilus-2.24.1.ebuild: alpha arm hppa ia64 ppc ppc64 sh sparc x86-fbsd -> exempi, https://bugs.gentoo.org/show_bug.cgi?id=242812 So that makes: >=media-gfx/eog-2.24.0 >=gnome-base/nautilus/nautilus-2.24.0 please don't forgot to rekeyword those when you're done with exempi, thanks in advance.
It needs valgrind :-( which doesnt work on bsd, can't proceed unless this can be made optional. masking test useflag could be an option but then I don't know what will happen if I run with FEATURES=test. Moreover it needs a patch to build because it assumes iconv functions are in the libc...
Created attachment 170012 [details, diff] patch needed to build on bsd With that patch it builds on bsd, beware of the dep on iconv.m4. Tests aren't ran if we don't have valgrind so I assume it's fine to mask the flag.
Created attachment 170359 [details] media-libs:exempi-2.0.2:20081030-223630.log [hppa] The build system depends on valgrind automagically, so I don't see any need for that DEPEND: checking for valgrind... no checking valgrind/memcheck.h usability... no checking valgrind/memcheck.h presence... no checking for valgrind/memcheck.h... no configure: WARNING: Valgrind is missing. checks will run without As the build log shows, however, it still does need something fixed. :)
That's with dev-libs/boost-1.35.0-r2 btw
Regardless of the test suite, media-libs/exempi builds and works fine (judging from the File -> Properties dialog in media-gfx/eog) so the only thing that stands in the way of keywording is the unnecessary DEPEND on valgrind - why don't we change that DEPEND into an einfo in pkg_setup that suggests that valgrind may help on certain platforms.
(In reply to comment #6) > Regardless of the test suite, media-libs/exempi builds and works fine (judging > from the File -> Properties dialog in media-gfx/eog) so the only thing that > stands in the way of keywording is the unnecessary DEPEND on valgrind - why > don't we change that DEPEND into an einfo in pkg_setup that suggests that > valgrind may help on certain platforms. > valgrind only works on amd64,ppc,ppc64 and x86...therefore i'd say just p.use.mask exempi test use-flag
built on my efika, without the test keyword though as i didn't want to build valgrind or boost on it - seems to run fine with nautilus 2.24.1
(In reply to comment #7) > (In reply to comment #6) > > Regardless of the test suite, media-libs/exempi builds and works fine (judging > > from the File -> Properties dialog in media-gfx/eog) so the only thing that > > stands in the way of keywording is the unnecessary DEPEND on valgrind - why > > don't we change that DEPEND into an einfo in pkg_setup that suggests that > > valgrind may help on certain platforms. > > > valgrind only works on amd64,ppc,ppc64 and x86...therefore i'd say just > p.use.mask exempi test use-flag The test suite runs fine without valgrind...
(In reply to comment #7) > valgrind only works on amd64,ppc,ppc64 and x86...therefore i'd say just > p.use.mask exempi test use-flag So how about this: I mask `media-libs/exempi test' globally and re-CC arches that should then unmask it in their arch profiles... (Nobody else seems to want to do this simple thing.)
A week passes...
so I discussed this on irc with armin76 and as you say we can either p.use.mask but I don't like very much profile games or we can just drop valgrind from DEPEND with a comment in the ebuild. It would then be automagic but I believe this is not the only package doing this. Bringing this up to dev would probably be interesting. I'd vote for the "remove from DEPEND" solution for now. What would you say guys ?
For me the valgrind checks seem to stall or maybe I just need to wait more with no CPU usage going on... Anyhow, I need to package.use.mask exempi use flag for all the arches that haven't keyworded it yet, to be able to restore their arch keyword on various packages like eog and nautilus because otherwise their ~arch is broken (gnome-desktop-2.24 is in their ~arch, but eog/nautilus-2.24 not and the earlier versions don't build against gnome-desktop-2.24)
(In reply to comment #13) > For me the valgrind checks seem to stall or maybe I just need to wait more with > no CPU usage going on... > > Anyhow, I need to package.use.mask exempi use flag for all the arches that > haven't keyworded it yet Why don't you just drop the valgrind dependency and allow us to keyword exempi?
(In reply to comment #14) > Why don't you just drop the valgrind dependency and allow us to keyword exempi? Or you could take tests as an automated test to find out possible problems, but not take them as definite if it is investigated and proven they are really not a problem in the application themselves, and not hold back keywording because of that. I'll remove the DEPEND in a moment, please proceed.
Thanks, marked ~hppa now (and eog).
~alpha/~ia64 done. need to debug why it sigbuses on sparc when opening a png file. hppa: you're missing your keyword on nautilus, unless its done on purpose
(In reply to comment #17) > hppa: you're missing your keyword on nautilus, unless its done on purpose Thanks. All done now, then.
~sparc done
Marked ~ppc/~ppc64.
I did this for arm/sh already...
ppc64 done
(In reply to comment #2) > It needs valgrind :-( which doesnt work on bsd, can't proceed unless this can > be made optional. masking test useflag could be an option but then I don't know > what will happen if I run with FEATURES=test. Moreover it needs a patch to > build because it assumes iconv functions are in the libc... > we can skip tests for bsd writing a custom src_test or look into patching out tests needing valgrind for you. Could you report your iconv patch upstream please ? (or do you want me to do it ?)
package.use.mask for the test USE flag until valgrind dependency is needed should be fine, running FEATURES=test on a test USE masked ebuild is supposed to work out fine (if it doesn't it's a package manager bug), so I guess we should go for that for now.
Didn't want to close the bug, sorry.
I'll test patch with latest exempi asap.
(In reply to comment #3) > Created an attachment (id=170012) [details] > patch needed to build on bsd > > With that patch it builds on bsd, beware of the dep on iconv.m4. > Tests aren't ran if we don't have valgrind so I assume it's fine to mask the > flag. > Thanks. + 17 Feb 2010; Samuli Suominen <ssuominen@gentoo.org> exempi-2.1.1.ebuild, + +files/exempi-2.1.1-iconv.patch: + Fix missing -liconv with ~x86-fbsd and keyword wrt #242812. Let me know if you get any problems, pushing this to upstream now.
https://bugs.freedesktop.org/show_bug.cgi?id=26602