the mass rebuild and i586 rpms?

[Milos Jakubicek]

Hi,

On 22.10.2009 19:29, Quentin Armitage wrote:
> 1. Is the script that is run and produces the output at
> http://jkeating.fedorapeople.org/needed-f12-rebuilds.html actually the
> script referred to at the bottom of that page
> (https://fedorahosted.org/rel-eng/browser/scripts) ? The reason I ask is
> that when I run the script, I get
>
> Included Koji instances:
> http://koji.fedoraproject.org/kojihub
> http://sparc.koji.fedoraproject.org/kojihub
> http://s390.koji.fedoraproject.org/kojihub
> http://arm.koji.fedoraproject.org/kojihub
>
> whereas the posted output only has
> Included Koji instances:
> http://koji.fedoraproject.org/kojihub
> http://sparc.koji.fedoraproject.org/kojihub

It is, but Jesse has disabled the other Koji hubs because sometimes they 
just time out, unfortunately.

> 2. If the script is run against just koji.fedoraproject,org/kojihub
> (i.e. without the sub arches), it says 185 packages need
> rebuilding (instead of the 175 listed in the report); the following
> 10 packages are omitted when the sparc koji hub is also included:
> gmpc
> HippoDraw
> itcl
> latex2rtf
> prtconf
> PyKDE
> python-igraph
> silo
> spicebird
> xorg-x11-drv-sunffb
>
> This is caused by line 117 of the script:
>          unbuilt = unbuilt&  unbuiltnew
> so if a package needs to be rebuilt on the primary arch, but not on the
> (in this case sparc) secondary arch, then it is dropped from needing to
> be rebuilt

Yes, that's how I did it -- my primary goal was to clear the list off 
secondary-arch-only stuff. There might be of course some cases like if 
somebody rebuilds a package only for a secondary arch but not for the 
primary one, but I don't think this is much a problem (there won't be 
many compared to the -- increasing -- number of secondary-arch-only 
stuff which we won't need).

  (it appears that a package will only be listed if it needs
> to be rebuild on every arch).

No, the package appears if there is no build after the specified date in 
any of the archs (to be clear: as soon as the package is built in at 
least one arch, it will get off the list).

  There are several circumstances where this
> can happen (with the 10 missing packages listed):
>
> Built on sub arch but failed on primary arch
> ============================================
> gmpc - 0.18.0-1 build on sparc after epoch but 0.18.0-2 failed on koji
> HippoDraw
> itcl
> latex2rtf
> python-igraph

Yes, those are not caught by this script now and should be rebuilt in 
primary arch as well of course.

> Not a primary arch package (should the package be blocked in the primary
> arch kojihub?)
> ==========================
> prtconf
> silo
> xorg-x11-drv-sunffb

There are much more of them! I don't know whether it is possible to 
block a package in a single Koji hub and if our infrastructure team is 
willing to go in this way -- Jesse?

> Blocked on secondary arch (so not included in unbuiltnew)
> =========================
> spicebird

Should be probably blocked in all hubs too. The blocking mechanism 
definitely doesn't serve instead of ExcludeArch, am I right?

> Built on sub arch but not submitted for rebuild on primary arch
> ================================================================
> PyKDE

Should be rebuilt (I just started the build).

> Package does not exist in secondary arch (no example)
> =====================================================
>
> Would it be more relevant to list what needs to be rebuild separately
> for each arch (but see point 3 below)?
> 3. So far as I can see, there have not been mass rebuilds on the
> secondary arches, so is it relevant to search them for successful builds
> since the epoch? If it is relevant, they would appear to have different
> epochs in any case.

Well, when I got to modifying the script (about half a year ago), the 
main problem was that there was too much noise consisting in 
secondary-arch-only packages. At that time there were more than 100 of 
such builds which is quite a lot.

Also, secondary archs (re)builds are completely in the hands of 
secondary arch maintainers, they're not bound to the primary archs mass 
rebuilds.

> 5. I have looked at the 185 packages that have not been rebuilt, and the
> reasons fall into the following categories (details for each package are
> listed later):
> 1. Not submitted for rebuild (65)

Yeah, there were some problems during the mass rebuild, IIRC (esp. with 
packages starting with o*, p* and maybe others). They should be 
definitely rebuilt. Looking at your lists, when rebuilding packages you 
should be aware of:

1) secondary-arch-only packages (xorg-x11-drv-sun*)
2) dead packages not blocked in Koji (a package is dead iff there is a 
dead.package file in the CVS; if it is then not blocked in Koji, please 
report to Jesse).
3) packages not built yet because they've just passed the review.

> I have made some changes to need-rebuild.py to produce some of the
> information above, and am happy to provide them if they are of any
> interest.

Great! The more people get involved, the better:)

Regards,
Milos