Marc Espie wrote: > On Sat, Oct 20, 2012 at 05:03:22PM +0200, Christian Weisgerber wrote: > > Nigel Taylor: > > > > > The last mono I built was 8 Oct on amd64, 9 Oct on i386, didn't work > > > first time as I recall on amd64, that was using dpb. When I hit a > > > failure, I just use make, once I have finished with the dpb building, > > > which continues from the point of failure, I may have rebooted before > > > running the make. Never hit a mono recently where I couldn't build it, > > > after one or more retries. > > > > Right. Mono has a certain probability of building, i.e., sometimes > > it builds and sometimes it doesn't. THIS IS NOT ACCEPTABLE. > > > > Previously, when producing amd64 snapshots, I restarted the mono > > build as often as necessary to get it built, but I have stopped > > doing that. Sometimes mono packages will show up in a snapshot, > > but mostly they won't. > > > > Mono is broken. If people need it, they better fix it. Retrying > > builds until one succeeds does not constitute a fix. Build environment > > voodoo (dpb, no dpb, reboot before building) does not constitute a > > fix. > > I'm sure the reliability of mono builds has NOTHING to do with not using > dpb, and a lot to do with not doing anything significant on the same > machine concurrently. > > There is a race condition in that build process. > > It tends not to hit it if you're not doing anything else at all on the same > machine at the same time.
I'm not using mono, but brief look at FreeBSD port shows at least one threads related patch which is not in our port. http://www.freebsd.org/cgi/cvsweb.cgi/ports/lang/mono/files/patch-mono_metadata_appdomain.c https://bugzilla.novell.com/show_bug.cgi?id=528830 Not sure about mono build process, but if mono builds its parts by itself, threading issues may impact on this. Alexey
