[RFC] autopinger

[RFC] autopinger

[Giovanni Bajo]

Hello,

I'm shortly sending a first test of my autopinger script for pending patches to
gcc-patches@. Any comments are welcome. The plan would be running it weekly.

Thanks,
Giovanni Bajo

Re: [RFC] autopinger

[Ben Elliston]

"Giovanni Bajo" <giovannibajo@libero.it> writes:

> I'm shortly sending a first test of my autopinger script for pending
> patches to gcc-patches@. Any comments are welcome. The plan would be
> running it weekly.

What does it do, exactly?

Ben

Re: [RFC] autopinger

[Ranjit Mathew]

Giovanni Bajo wrote:

> Hello,
> 
> I'm shortly sending a first test of my autopinger script for pending patches to
> gcc-patches@. Any comments are welcome. The plan would be running it weekly.

I just saw the first such mail and I have a quick comment: it
might probably be better for the maintainers if the mail
can be split into separate ones for each area.

Ranjit.

-- 
Ranjit Mathew          Email: rmathew AT gmail DOT com

Bangalore, INDIA.      Web: http://ranjitmathew.tripod.com/

Re: [RFC] autopinger

[Giovanni Bajo]

Ben Elliston wrote:

>> I'm shortly sending a first test of my autopinger script for pending
>> patches to gcc-patches@. Any comments are welcome. The plan would be
>> running it weekly.
>
> What does it do, exactly?

It goes through Bugzilla looking for bugs with the keyword "patch" set, and
with a comment containing a link to the gcc-patches (or java-patches) mailing
list. Bugs matching this criteria are considered with a patch pending for
review. It generates a mail with an attacchment containing information on such
bugs, ordered by component, including PR number, PR title, Patch author, date
of patch, link to gcc-patches, link to PR in Bugzilla.

To "register" a posted patch to the automatic pinger, go to the corresponding
Bugzilla entry (if any -- or you can open one), add the keyword "patch", and a
comment with a link to the patch. Notice that this is a standard convention
already in practice among maintainers and Bugmasters for a long time already.
Patchpinger just relies on this because it is already being done.

To "deregister" a patch which was rejected by the review, just remove the
keyword "patch" from the bug.

The automatic pinger was designed to *not* force maintainers to learn/use a new
way of handling patches. It relies on widely used conventions, and does not
force people to do special things they never did before. In other words, it is
mostly *transparent* to users.

Giovanni Bajo

Re: [RFC] autopinger

[Giovanni Bajo]

Ranjit Mathew wrote:

> I just saw the first such mail and I have a quick comment: it
> might probably be better for the maintainers if the mail
> can be split into separate ones for each area.

Are you suggesting one different mail for components? This would generate a
flood of a dozen of mails. I do not think people would like such a flood, I
heard people are already annoyed because this pinger will generate one more
mail per week (!).

Or did you mean different attacchments within the same mail?

Giovanni Bajo

Re: [RFC] autopinger

[Ian Lance Taylor]

"Giovanni Bajo" <giovannibajo@libero.it> writes:

> I'm shortly sending a first test of my autopinger script for pending patches to
> gcc-patches@. Any comments are welcome. The plan would be running it weekly.

For each PR, if the PR is assigned to somebody, I recommend that you
mention that person in the list.  Perhaps the patch pinger should CC
those people; perhaps not.

I also think the message should list the priority, the target
milestone, and the "known to work" and "known to fail" fields.  That
will let people more easily prioritize the information in the message
without having to go look at the PR.

Thanks for doing this.

Ian

Re: [RFC] autopinger

[Giovanni Bajo]

Ian Lance Taylor wrote:

> For each PR, if the PR is assigned to somebody, I recommend that you
> mention that person in the list.  Perhaps the patch pinger should CC
> those people; perhaps not.

The pinger attempts to find out the "author" of the patch. The first choice is
the author of the mail to gcc-patches; the second choice is the assigned field
in Bugzilla. As a matter of fact, right now I only post patches which have an
associated mail to gcc-patches, so the second choice is never taken. It would
become useful when/if I am asked to make it ping also patches attacched to
Bugzilla, which some people don't like because it violates our rules that each
patch should be posted to gcc-patches[1].

This said, do you believe that explicitally.listing the assignee of the bug
together with the "author" of the patch (mail sender) is still useful? What
would the rationale be?

> I also think the message should list the priority, the target
> milestone, and the "known to work" and "known to fail" fields.  That
> will let people more easily prioritize the information in the message
> without having to go look at the PR.

Sure, I will imlement these. I'm assuming that for "priority" you actually mean
"severity", since we are not using the priority field at this point.

Thanks for the feedback.

Giovanni Bajo


[1] I just had an idea. If the pinger finds a patch in Bugzilla which is not
sent to gcc-patches, it could generate a ping mail to the author of the patch,
remembering him to post it to gcc-patches, until he does, or the attacchment is
marked as obsolete (if the patch is invalid).

Re: [RFC] autopinger

[Ian Lance Taylor]

"Giovanni Bajo" <giovannibajo@libero.it> writes:

> > For each PR, if the PR is assigned to somebody, I recommend that you
> > mention that person in the list.  Perhaps the patch pinger should CC
> > those people; perhaps not.
> 
> The pinger attempts to find out the "author" of the patch. The first choice is
> the author of the mail to gcc-patches; the second choice is the assigned field
> in Bugzilla. As a matter of fact, right now I only post patches which have an
> associated mail to gcc-patches, so the second choice is never taken. It would
> become useful when/if I am asked to make it ping also patches attacched to
> Bugzilla, which some people don't like because it violates our rules that each
> patch should be posted to gcc-patches[1].
> 
> This said, do you believe that explicitally.listing the assignee of the bug
> together with the "author" of the patch (mail sender) is still useful? What
> would the rationale be?

I do think so, because the person to whom the PR is assigned, if
anyone, is some gcc maintainer who has some responsibility for
resolving the problem.  If that person is not the same as the author
of the patch, then he or she should be looking at the patch to accept
or reject it, or should reassign the bug to some other gcc maintainer.

> Sure, I will imlement these. I'm assuming that for "priority" you actually mean
> "severity", since we are not using the priority field at this point.

Indeed.

Thanks.

Ian

Re: [RFC] autopinger

[Giovanni Bajo]

Ian Lance Taylor wrote:

> I do think so, because the person to whom the PR is assigned, if
> anyone, is some gcc maintainer who has some responsibility for
> resolving the problem.  If that person is not the same as the author
> of the patch, then he or she should be looking at the patch to accept
> or reject it, or should reassign the bug to some other gcc maintainer.

OK, I'll report also the assignee iff he's the not the patch author.

Giovanni Bajo