Proposal: Better force-tag

Adam Jackson ajax at redhat.com
Mon Sep 15 13:35:45 UTC 2008


On Sun, 2008-09-14 at 12:37 -0400, Warren Togami wrote:
> Steve Grubb wrote:
> > On Sunday 14 September 2008 10:16:03 Warren Togami wrote:
> >>> Just removing stuff does not solve problems you have to implement a
> >>> better alternative first (and not rely on a worse one).
> >> Since nothing stops the user from using TAG_OPTS=-F, it is ineffective
> >> of us to remove force-tag.  How about an improved force-tag instead?
> >>
> >> Make force-tag use koji to check if a build for that N-V-R is already
> >> complete.  If it is complete, then refuse to tag again with the same N-V-R.
> > 
> > Yes, this sounds exactly like what should have been done. We should not have 
> > disabled a feature that many people legitimately use until we had a tested 
> > solution that does not interrupt the work flow. This is needless arguing with 
> > no tangible benefits for anyone. I consider the srpm to be the canonical 
> > source of the build. That is the only granularity people can access via any 
> > mirror.
> 
> The goal of preventing force-tag is to better guarantee that .src.rpm 
> comtents match tags.  We can achieve this by having force-tag fail if 
> the buildsystem says that E:N-V-R was either already built or is 
> currently building.

Except, as mentioned elsewhere in the thread, the goal is really "srpm
contents are reconstructable".  Which is exactly equivalent to saving
the Entries file from the CVS checkout as part of the build output.

A patch to implement this exists.

- ajax
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
URL: <http://listman.redhat.com/archives/fedora-devel-list/attachments/20080915/dd0faad2/attachment.sig>


More information about the fedora-devel-list mailing list