Nicolas Chauvet píše v So 27. 10. 2012 v 12:40 +0200: > Yesterday. > Le 27 oct. 2012 10:43, "Julian Sikorski" <beleg...@gmail.com> a écrit : > > > > W dniu 11.10.2012 07:28, Julian Sikorski pisze: > > > W dniu 11.10.2012 00:27, Nicolas Chauvet pisze: > > >> 2012/10/10 Julian Sikorski < > belegdol-re5jqeeqqe8avxtiumwx3w-xmd5yjdbdmrexy1tmh2...@public.gmane.org>: > > >>> W dniu 2012-10-10 12:22, Leigh Scott pisze: > > >>>> > > >>>> Hi Julian, > > >>>> > > >>>> > > >>>>> If I could somehow force only one build at a > > >>>>> time, this could help. Otherwise we might need to bump the builder > > >>>>> memory - how much does it have now? > > >> That's the point, the builder is lacking memory, but only during link > > >> time of a given build. > > >> I've reproduced the problem on a similar builder. I will try to > > >> reproduce using a f18 target. > > >> > > >> Please do not submit any mame jobs for now. > > >> Thx > > >> > > >> Nicolas (kwizart) > > >> > > > Sure, no problem. If the memory is the only issue, I can consider > > > donating some. Let me know what your tests show and what type of memory > > > is needed. > > > > > > Julian > > > > > Where do we stand on this? If RPM Fusion is no longer able to supply > > up-to-date mame packages, I need to look for another solution. > > > > Regards, > > Julian > > Hi Julian, > > I was only able to test a f18 target > Yesterday. And it has failed the same as f19. > > There are actions ongoing to solve this from the infra side. > > Btw. Is there a way to lower the memory usage needed at link time until the > problem is solved ?
the trick used in Fedora on s390/ppc/arm for some packages (webkit, ...) was to switch from -g to -g1 (generates less debuginfos) and use "-Wl,--no-keep-memory -Wl,--reduce-memory-overheads" options for the linker Dan