[patch] gdb/CONTRIBUTE update

[patch] gdb/CONTRIBUTE update

[Jan Kratochvil]

Hi Eli,

there is one s/conclusion/inclusion/ fix reviewed by Phil.

Otherwise TBH I did not well understand the "distribute the change yourself"
encouragement, it seems to me like a suggestion one should keep 3rd party
patches, these are discouraged, patches should be integrated to FSF GDB.

And added the new paragraph as people send patch and as they expect the GDB
maintainers will grab it and handle it on their own many patches get forgotten
that way as there is no patch tracking in place (yet).

Just such an idea, np if people consider it OK as is.


Thanks,
Jan


gdb/
2013-02-21  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* CONTRIBUTE: Fix type to inclusion.  Extend patches distribution
	description.  Add new paragraph on patch ownership.

diff --git a/gdb/CONTRIBUTE b/gdb/CONTRIBUTE
index cdec246..67af0ab 100644
--- a/gdb/CONTRIBUTE
+++ b/gdb/CONTRIBUTE
@@ -10,8 +10,9 @@ included in the GDB distribution, the GDB web pages also contain much
 information.
 
 You may also want to submit your change so that can be considered for
-conclusion in a future version of GDB (see below).  Regardless, we
-encourage you to distribute the change yourself.
+inclusion in a future version of GDB (see below).  Regardless, we
+encourage you to distribute the change yourself until it gets integrated
+into FSF GDB mainline sources.
 
 If you don't feel up to hacking GDB, there are still plenty of ways to
 help!  You can answer questions on the mailing lists, write
@@ -86,6 +87,12 @@ o	Submitting Patches
 	CVS repository, see the Anonymous read-only CVS access and
 	Read-write CVS access page.
 
+	Patch from a contributor needs a review with approval from
+	maintainer.  Still you drive the review process and inclusion
+	process of the patch.  If there is no reply in a week send a new
+	mail (not reply) with PING in its subject.  Occasionally even
+	a PING^2 mail may be needed in another week of no replise.
+
 --
 
 Supplemental information for GDB:

Re: [patch] gdb/CONTRIBUTE update

[Tom Tromey]

>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:

Jan> +	Patch from a contributor needs a review with approval from

I would write "A patch from..."

Jan> +	mail (not reply) with PING in its subject.  Occasionally even
Jan> +	a PING^2 mail may be needed in another week of no replise.

Typo, "replies".

Tom

Re: [patch] gdb/CONTRIBUTE update

[Yao Qi]

On 02/22/2013 04:26 AM, Jan Kratochvil wrote:
> +	Patch from a contributor needs a review with approval from
> +	maintainer.  Still you drive the review process and inclusion
> +	process of the patch.  If there is no reply in a week send a new
> +	mail (not reply) with PING in its subject.  Occasionally even
> +	a PING^2  mail may be needed in another week of no replise.
> +

Jan,
why do we have to post the patch again in the new mail instead of a 
reply to remind maintainers to review?  If the patch is still applied 
clearly, I don't see the benefits of doing that.

-- 
Yao (齐尧)

Re: [patch] gdb/CONTRIBUTE update

[Jan Kratochvil]

On Tue, 26 Feb 2013 10:31:06 +0100, Yao Qi wrote:
> On 02/22/2013 04:26 AM, Jan Kratochvil wrote:
> >+	Patch from a contributor needs a review with approval from
> >+	maintainer.  Still you drive the review process and inclusion
> >+	process of the patch.  If there is no reply in a week send a new
> >+	mail (not reply) with PING in its subject.  Occasionally even
> >+	a PING^2  mail may be needed in another week of no replise.
> >+
> 
> Jan,
> why do we have to post the patch again in the new mail instead of a
> reply to remind maintainers to review?  If the patch is still
> applied clearly, I don't see the benefits of doing that.

With a reply mail users (reviewers) which use sorting of mail folder by
threads (in Mutt 'o' 't') get the PING mail put under the original mail which
is far in the past and the PING mail gets hidden+forgotten again due to it.

This even is not an idea of mine, it was concluded on some GNU Tools Cauldron.

I do not speak here whether the patch itself should be included in the PING
mail again or not, I do not find that important and it is also not expressed
explicitly in the proposed paragraph above.


Thanks,
Jan

Re: [patch] gdb/CONTRIBUTE update

[Yao Qi]

On 02/26/2013 05:38 PM, Jan Kratochvil wrote:
> This even is not an idea of mine, it was concluded on some GNU Tools Cauldron.
>

If you meant Cauldron last year, I can't recall any discussions on this.

> I do not speak here whether the patch itself should be included in the PING
> mail again or not, I do not find that important and it is also not expressed
> explicitly in the proposed paragraph above.

OK.  The subject is prefixed with "PING".  What should we put in the 
body of the mail?  The url of the mail archive or something else?

-- 
Yao (齐尧)

Re: [patch] gdb/CONTRIBUTE update

[Pedro Alves]

On 02/26/2013 09:38 AM, Jan Kratochvil wrote:
> On Tue, 26 Feb 2013 10:31:06 +0100, Yao Qi wrote:
>> On 02/22/2013 04:26 AM, Jan Kratochvil wrote:
>>> +	Patch from a contributor needs a review with approval from
>>> +	maintainer.  Still you drive the review process and inclusion
>>> +	process of the patch.  If there is no reply in a week send a new
>>> +	mail (not reply) with PING in its subject.  Occasionally even
>>> +	a PING^2  mail may be needed in another week of no replise.
>>> +
>>
>> Jan,
>> why do we have to post the patch again in the new mail instead of a
>> reply to remind maintainers to review?  If the patch is still
>> applied clearly, I don't see the benefits of doing that.
> 
> With a reply mail users (reviewers) which use sorting of mail folder by
> threads (in Mutt 'o' 't') get the PING mail put under the original mail which
> is far in the past and the PING mail gets hidden+forgotten again due to it.
> 
> This even is not an idea of mine, it was concluded on some GNU Tools Cauldron.

Do you like it?  I don't.

Interesting.  I had seen some pings like that on the gcc list, but I never
understood why people preferred them that way.

One consequence of that model I've often seen in the gcc list (and even
there only a few people tend to ping that way) is occasionally we see
the reviewer OKing in reply to the ping email, instead of replying to
the original submission, which obviously happens because it was quicker
to the reviewer to just hit reply to the ping, read the patch in the
browser (following the url), and hit send.  I don't like this because
it makes archaeology harder, and, people who are tracking the original
thread may not notice the patch had been approved or further discussion had
happened on another thread.

If someone is pinging me, I prefer a reply to the original thread.  I'm
currently using Thunderbird, and it sorts threads by most recent reply,
not by original post date (which I find a more sensible default - why
would one want to sort by original post date by default?).  A separate
email with an url means more work for me, as I then have to go look up the
thread in Thunderbird that corresponds to that url, instead of having it
already handy (as when the ping is in the same thread, and thus displayed
as a child of the original patch).  That's just a silly indirection out of
email to then reply back through email.  It's a little more work for the
pinger as well, who has to look up an url to put in the email.

-- 
Pedro Alves

Re: [patch] gdb/CONTRIBUTE update

[Jan Kratochvil]

On Tue, 26 Feb 2013 12:16:38 +0100, Pedro Alves wrote:
> On 02/26/2013 09:38 AM, Jan Kratochvil wrote:
> > With a reply mail users (reviewers) which use sorting of mail folder by
> > threads (in Mutt 'o' 't') get the PING mail put under the original mail which
> > is far in the past and the PING mail gets hidden+forgotten again due to it.
> > 
> > This even is not an idea of mine, it was concluded on some GNU Tools Cauldron.
> 
> Do you like it?  I don't.

I do.


> Interesting.  I had seen some pings like that on the gcc list, but I never
> understood why people preferred them that way.

The discussion at Cauldron was primarily for GCC.


> I'm currently using Thunderbird, and it sorts threads by most recent reply,
> not by original post date

I do not see such option in Mutt.  Mutt could be sure extended but as I can
imagine it is also not perfect as old but still discussed threads clutter new
posts.  Besides that there are other mailers like VM/Gnus which also may not
have thre thread-by-most-recent-reply sorting, not sure.


> A separate email with an url means more work for me, as I then have to go
> look up the thread in Thunderbird that corresponds to that url,

This is why I started to include also Message-IDs besides the URLs. Posting
just the Message-IDs is unfortunately insufficient.

Looking up the URL from mailer/Mutt is also a pain, I wanted to code some
script for Message-ID -> URL mapping but as sourceware server should be
replaced ASAP which will replace MHonArc by PiperMail I have not yet coded it;
unfortunately the ASAP replace takes already some years.


Unfortunately whatever method we find out it always is a pain, the reviewing
process should be supported by projects like Gerrit.  Until that is
tried/established the goal is just to find the least pain.


Jan

Re: [patch] gdb/CONTRIBUTE update

[Jan Kratochvil]

On Tue, 26 Feb 2013 10:50:59 +0100, Yao Qi wrote:
> On 02/26/2013 05:38 PM, Jan Kratochvil wrote:
> >This even is not an idea of mine, it was concluded on some GNU Tools Cauldron.
> 
> If you meant Cauldron last year, I can't recall any discussions on this.

IIRC it was there in Prague, it was just a small BoF, not too many people
there.


> >I do not speak here whether the patch itself should be included in the PING
> >mail again or not, I do not find that important and it is also not expressed
> >explicitly in the proposed paragraph above.
> 
> OK.  The subject is prefixed with "PING".  What should we put in the
> body of the mail?  The url of the mail archive or something else?

Any or IMO best all of:
	Patch repost.
	URL of the original message.
	Message-ID of the original message.

This is all the subject of discussion as can be seen.


Jan

Re: [patch] gdb/CONTRIBUTE update

[Tom Tromey]

>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:

Jan>  Besides that there are other mailers like VM/Gnus which also may
Jan> not have thre thread-by-most-recent-reply sorting, not sure.

gnus works fine here.
It doesn't sort this way (at least not for me, I didn't look to see if
it can), but it marks new messages regardless of where they appear in
the sorting.

I suppose I have a slight preference for posting as a follow-up.
But I will cope with anything.

Jan> Unfortunately whatever method we find out it always is a pain, the
Jan> reviewing process should be supported by projects like Gerrit.

I think we owe it to patch submitters to do this.
Pings don't really seem to be working.

Tom

Re: [patch] gdb/CONTRIBUTE update

[Tom Tromey]

Tom> I think we owe it to patch submitters to do this.
Tom> Pings don't really seem to be working.

Jan> I did not get what should the submitters do if even pings do not work.

I was thinking of automated pings.  Ideally there's really no need for
pings if the list of unreviewed patches is maintained by a tool rather
than by each of us in our own inbox -- but a weekly automated report or
something wouldn't hurt.

Tom

Re: [patch] gdb/CONTRIBUTE update

[Jan Kratochvil]

On Fri, 01 Mar 2013 17:27:07 +0100, Tom Tromey wrote:
> I was thinking of automated pings.  Ideally there's really no need for
> pings if the list of unreviewed patches is maintained by a tool rather
> than by each of us in our own inbox -- but a weekly automated report or
> something wouldn't hurt.

That is the whole problem  - how to keep the list of pending patches.

I maintain a list of my own patches I submitted.  Another problem is that
older patches sometimes even reviewed no longer get a response by their
submitter so there should be also some response timeouts.


Jan

Re: [patch] gdb/CONTRIBUTE update

[Tom Tromey]

Jan> That is the whole problem  - how to keep the list of pending patches.

Jan> I maintain a list of my own patches I submitted.  Another problem is that
Jan> older patches sometimes even reviewed no longer get a response by their
Jan> submitter so there should be also some response timeouts.

I think bugzilla would be better.

* User files patch in bugzilla, makes sure to tag the attachment as a patch
* Reviews can be done in bugzilla (or email?  I forget if that is
  enabled on sourceware, but if not perhaps we could set it up)
* A reviewer would put the bug into WAITING; we can randomly do queries
  for this and close all WAITING bugs older than a year or two
* We could have a job anywhere to query bugzilla and send out weekly
  ping reminders


To be clear, I don't object to your CONTRIBUTE change.
I have a suspicion that adding more work for the contributors is going
to mean more patches being dropped (by them), but time will tell.

Tom