Meeting log from last week

Michael Schwendt mschwendt.tmp0701.nospam at arcor.de
Mon Mar 5 16:43:22 UTC 2007


On Sat, 03 Mar 2007 18:04:42 +0100, Thorsten Leemhuis wrote:

> Hi,
> 
> just FYI, I uploaded the meeting log from last week to the wiki:
> 
> http://www.fedoraproject.org/wiki/EPEL/Meetings/20070225

Two corrections here:

> 00:20 <         thl> | the current push scripts also don't handle testing nicely

They don't handle it at all. More important, it's not implemented in the
buildsys, so there's no support for it to start with. Consider that
everything that enters "needsign" is available to subsequent build-jobs
immediately. No separate build targets like "epel4" and "testing/epel4"
are available.

Writing a simple build-results mover should not be difficult, even without
a package database that would keep track of where a build-job's packages
have been pushed to. Similar to how RepoPrune does it, one can find a set
of source+binary rpms for a given pkg %{name} based on the "Source RPM"
header and then move/copy everything to a different target repo.

The more recent push scripts can release single build-jobs, ignore the
rest of the needsign queue, and at least provide a rough way to include
the needsign queue in the repoclosure check (*with* mail reports to the
pkg owners). Missing is the integration of the form "only push all pkgs
that passed the test and none that were built after/during the test". With
a crowded needsign queue, one doesn't want to cut'n'paste many package
names, although that is possible. And, of course, when something is
reported as broken, somebody would need to investigate, exclude packages
and rerun time-consuming tests while packagers submit even more build
jobs. With available code and no optimisations, one can easily see that
running this automatically and frequently would occupy a dedicated server:

$ time pushscript/RCNeedsign.py Extras development
Copying build results to temporary working directory:
  gkrellm-moon-0.6-2.fc7
  gkrellm-sun-1.0.0-2.fc7
  gquilt-0.20-1.fc7
  qps-1.9.19-0.1.a.fc7
  qt4-4.2.2-7.fc7
  sysprof-kmod-1.0.8-1.2.6.20_1.2962.fc7
Installing into temporary repository:
  gkrellm-moon-0.6-2.fc7
  gkrellm-sun-1.0.0-2.fc7
  gquilt-0.20-1.fc7
  qps-1.9.19-0.1.a.fc7
  qt4-4.2.2-7.fc7
  sysprof-kmod-1.0.8-1.2.6.20_1.2962.fc7
Creating repository /tmp/tmp-repo-Lbc9uW/development/ppc
Creating repository /tmp/tmp-repo-Lbc9uW/development/x86_64
Creating repository /tmp/tmp-repo-Lbc9uW/development/i386
Running /srv/extras-push/work/extras-repoclosure/rc-run.py --mail=owners --needsign=file:///tmp/tmp-repo-Lbc9uW/%s/%s/ development
Trying to acquire lock file,
waiting for any active job to terminate: OK
development (ppc)
development (x86_64)
development (i386)

real    12m55.152s
user    7m20.028s
sys     0m47.181s


> 00:22 <    dgilmore> | pretty much only one person pushes extras
> 00:22 <       nirik> | I could possibly also do sign/pushes on a schedule. ie, every morning at some time, etc.
> 00:22 <         thl> | having three or four signers that can push sounds enough to me
> 00:23 <       nirik> | yeah, mschwent?
> 00:23 <         thl> | nirik, I think so

Ville and me push regularly. Warren probably only if we don't push often
enough :) or if the availability of important updates matches his
time-zone.




More information about the epel-devel-list mailing list