On Fri, Jul 31, 2009 at 3:09 PM, Peter FELECAN<[email protected]> wrote: > "Maciej (Matchek) Blizinski" <[email protected]> writes: > >> On Fri, Jul 31, 2009 at 11:42 AM, Peter FELECAN<[email protected]> wrote: >>> "Maciej (Matchek) Blizinski" <[email protected]> writes: >>> >>>> On Fri, Jul 31, 2009 at 11:31 AM, Peter FELECAN<[email protected]> >>>> wrote: >>>>> "Maciej (Matchek) Blizinski" <[email protected]> writes: >>>>> >>>>>> 2. Avoid building certain blacklisted packages. Optional: schedule >>>>>> periodic or nightly builds for those packages. >>>>> >>>>> -1 as it defeats the purpose; in addition, there is no night for an >>>>> international community as ours... >>>> >>>> Automatic builds for some packages are better than automatic builds >>>> for no packages, so it doesn't defeat the purpose completely. Also, we >>>> could do actual measurements: how big is the impact on the >>>> performance? >>> >>> Maybe. But not on the effective build stack, please. >> >> Where are large packages currently built? I thought that all packages >> must be built on the buildfarm. Next time there's an on-going >> compilation of a large package, we can do the measurements. > > Some weeks ago we had I/O issues when gcc3, OpenOffice and Firefox where > built in the same period. What I'm saying, is *automatic build must be > done on a different hardware stack*
buildbot won't build more than one package on a single system. It might build more than on the same hardware, because it doesn't know about the VM layout. >> The GAR packages will be the main focus now, since they comprise about >> 91% builds (that I know of, that is, ~1083 out of ~1186). Once we're >> there, we can extend support to other build sources. Of course, it >> would make it easier to implement if a code repository was available >> to buildbot. > > Are you sure for the figures above? Do you want to argue about the figures? :-) I counted the packages I knew about: the number of subdirectories in mgar/pkg, plus mgar/pkg/cpan, plus number of packages listed on your maitainer page. Feel free to offer better figures! :-) Maciej _______________________________________________ maintainers mailing list [email protected] https://lists.opencsw.org/mailman/listinfo/maintainers
