Date: Sun, 23 Jul 2006 20:37:31 +0100 From: RW <list-freebsd-2004@morbius.sent.com> To: freebsd-ports@freebsd.org Subject: Re: portinstall breaks with -m "-j 4" Message-ID: <200607232037.32575.list-freebsd-2004@morbius.sent.com> In-Reply-To: <20060722211326.GB8547@soaustin.net> References: <44C1F68E.80601@gmx.de> <44C24549.4030703@gmx.de> <20060722211326.GB8547@soaustin.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Saturday 22 July 2006 22:13, Mark Linimon wrote: > > BTW, I apologize for this is not at all a portupgrade issue, but an issue > > of the ports system. > > It is an issue with individual ports -- actually not the "port" (e.g. > Makefile framework, pkg-*) but the individual applications (IIUC). > > > Well, at least the ports system itself should not be broken able to work > > with this. With larger ports I manage to reduce build times by 40% with > > distcc and a second machine. As far as I see it the number of ports > > breaking is rather low. > > Please feel free to suggest a framework (complete with regression test > framework) where the infrastructure code can "learn" which ports are safe. > I think it's going to be a harder problem than you think it is. Note that > "appears to work" and "can be shown to work under arbitrary build > circumstances for all users" are IMHO going to be two very different > classes of problem -- and the latter will need to be solved before it > can be used on the package-building cluster. It seem to me that virutally all the advantage could be obtained by passing -j just to the build stage, where portupgrade spend most of its time. In any case install is probably too IO-bound to benefit. The user could set say WITH_PARALLEL=4. The value could be passed down to the build if the port sets USE_PARALLEL=yes or the user sets WITH_PARALLEL_FORCE=yes.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200607232037.32575.list-freebsd-2004>