2014-07-15 17:16 GMT+02:00 Pierre-Yves Chibon <pingou@pingoured.fr>:

On Tue, Jul 15, 2014 at 05:04:26PM +0200, Thomas Spura wrote:
> 2014-07-15 16:35 GMT+02:00 Pierre-Yves Chibon <pingou@pingoured.fr>:
>

> New procedure (1)
> =================
>
> * packager opens a review-request on bugzilla
> * reviewer sets the fedora-review flag to ?
> * reviewer does the review
> * reviewer sets the fedora-review flag to +
> * packager goes to pkgdb2 to request new package

> A - precises package name
> A - precises package summary
> A - precises package branches
> A - precises link to review on bugzilla

> * requests added to the scm admin queue

> * cvsadmin checks the review (check reviewer is a packagerA^1)

> * cvsadmin approves the creation of the package in pkgdb
> * package creation is broadcasted on fedmsg
> * git adjusted automatically
>

> A^1 we could check this automatically by checking which comment mentions

> 'approved'
> and checking who set the fedora-review flag to +
>

> New procedure (2) A - Relies on fedmsg/bugzilla integration

> =================
>
> * packager opens a review-request on bugzilla
> * reviewer sets the fedora-review flag to ?
> * reviewer does the review
> * reviewer sets the fedora-review flag to +
> * automatically set fedora-cvs flag to ?
> * cvsadmin checks the review (check reviewer is a packager)
> * cvsadmin sets the fedora-cvs flag to +
> * flag change is broadcasted onto fedmsg
> * pkgdb automatically creates the package (w/ name and summary provided
> in the

> A review)

>
> How does pkgdb know the fas name of the packager? By the email of the
> reporter?

The email would be the way indeed. We do require packager to have the same email
in FAS and bugzilla and the admins get an hourly message when someone doesn't.

There have been (many?) cases, where the reporter didn't want to finish the review process and another one took over the review request in the same bug. So I wouldn't rely on this check as there are chances where the reviewer doesn't insisted on opening a new review request in such a case...

Or would it be possible to check all comments for SRPM/SPEC urls, so that all have been posted by the bug reporter?

> Maybe pkgdb2 could wait for the packager to approve the automatically
> filled values and then could kick off, when the packager confirms?

The idea is that summary will be updated via a cron taking the info from yum
anyway, so, if there was a typo in the summary, fixing it in the spec file will
fix it in pkgdb.

> This could then be merged with the next item below

> A

>
> * packager goes to pkgdb2 to request new branches

Merging validating name/summary with requesting new branches is possible, but it
will make things a little more complex as my idea is that this part of the
process would/could be used when requesting additional branches on an existing
package.

> Why is there a second check of an cvsadmin needed, when a new branch is
> created by the "Main Contact" of the new package?

That is the same check as the one we do already when requesting new branches

Which should happen automatically by the process script of the cvsadmin, isn't it?

If it is automatically there, it could also be done fully automatically without this second check by a cvsadmin.

> Shouldn't it be enough, to check above, if the review was sane with
> setting fedora-cvs to +?
>
> Another possibility would be to remove anything from above with
> "fedora-cvs" and only one check of an cvsadmin below would be required.

One advantage of keeping fedora-cvs is that it gives us the flexibility when/if
we miss or fedmsg drops a message.
But, we could consider it as well, I just seems to remember that Dennis wanted
to keep it in at the meeting.

>
> * requests added to the scm admin queue
> * cvsadmin approves the creation of the branches in pkgdb
> * branch creation is broadcasted on fedmsg
> * git adjusted automatically

Pierre

Tom