On Thu, Jun 14, 2012 at 11:05:33AM +0200, Jens Axboe wrote:
> On 06/14/2012 11:01 AM, Fengguang Wu wrote:
> > Hi Jens,
> >
> > On Thu, Jun 14, 2012 at 08:30:11AM +0200, Jens Axboe wrote:
> >> On 06/14/2012 01:03 AM, [email protected] wrote:
> >>> FYI, kernel build failed on
> >>>
> >>> tree: git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git multiqueue
> >>> head: e9a6aa8a0b7b93195774a9fcf551632bf3abcd97
> >>> commit: 67db7275fa4325d46853a4bfcb79acf32e48eca8 [1/3] multiqueue: a hodge podge of things
> >>
> >> The multiqueue branch is a private branch, it's known broken on many
> >> configs at the moment.
> >
> > OK. Sorry my script just blindly build tests every new/updated branches
> > in the tree. It does try to reduce noise level by remembering all warned
> > error messages. However if it still presents a problem to the private
> > branches, I can either explicitly blacklist "multiqueue", or let the
> > script automatically skip commits whose "^Signed-off-by:" is still missing.
>
> It's not a problem that you're building various branches, as long as you
> are expecting a bit of noise :-). In fact I applaud your effort on
> expanding the scope of building others branches.
>
> But it does raise the question of how do we best signal intermediate
> branches like this, so they can be avoided until they are ready. The
> blacklist doesn't seem like a great choice, since then I have to
> remember to remind you when it _is_ ok to build it.
>
> How about some arbitrary tag? Instead of checking a Signed-off-by, I
> could just add a Dont-Auto-Build or something like that. What do you
> think?
OK, I'll add a grep for "Dont-Auto-Build". If it's found in *any*
commit of the branch, the build script will skip the whole branch.
Does this make sense to you?
Thanks,
Fengguang
On 06/14/2012 12:02 PM, Fengguang Wu wrote:
> On Thu, Jun 14, 2012 at 11:05:33AM +0200, Jens Axboe wrote:
>> On 06/14/2012 11:01 AM, Fengguang Wu wrote:
>>> Hi Jens,
>>>
>>> On Thu, Jun 14, 2012 at 08:30:11AM +0200, Jens Axboe wrote:
>>>> On 06/14/2012 01:03 AM, [email protected] wrote:
>>>>> FYI, kernel build failed on
>>>>>
>>>>> tree: git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git multiqueue
>>>>> head: e9a6aa8a0b7b93195774a9fcf551632bf3abcd97
>>>>> commit: 67db7275fa4325d46853a4bfcb79acf32e48eca8 [1/3] multiqueue: a hodge podge of things
>>>>
>>>> The multiqueue branch is a private branch, it's known broken on many
>>>> configs at the moment.
>>>
>>> OK. Sorry my script just blindly build tests every new/updated branches
>>> in the tree. It does try to reduce noise level by remembering all warned
>>> error messages. However if it still presents a problem to the private
>>> branches, I can either explicitly blacklist "multiqueue", or let the
>>> script automatically skip commits whose "^Signed-off-by:" is still missing.
>>
>> It's not a problem that you're building various branches, as long as you
>> are expecting a bit of noise :-). In fact I applaud your effort on
>> expanding the scope of building others branches.
>>
>> But it does raise the question of how do we best signal intermediate
>> branches like this, so they can be avoided until they are ready. The
>> blacklist doesn't seem like a great choice, since then I have to
>> remember to remind you when it _is_ ok to build it.
>>
>> How about some arbitrary tag? Instead of checking a Signed-off-by, I
>> could just add a Dont-Auto-Build or something like that. What do you
>> think?
>
> OK, I'll add a grep for "Dont-Auto-Build". If it's found in *any*
> commit of the branch, the build script will skip the whole branch.
> Does this make sense to you?
Yeah that'll work for me. A description or tag might be more elegant,
but this is easier...
--
Jens Axboe
On Thu, Jun 14, 2012 at 12:45:38PM +0200, Jens Axboe wrote:
> On 06/14/2012 12:02 PM, Fengguang Wu wrote:
> > On Thu, Jun 14, 2012 at 11:05:33AM +0200, Jens Axboe wrote:
> >> On 06/14/2012 11:01 AM, Fengguang Wu wrote:
> >>> Hi Jens,
> >>>
> >>> On Thu, Jun 14, 2012 at 08:30:11AM +0200, Jens Axboe wrote:
> >>>> On 06/14/2012 01:03 AM, [email protected] wrote:
> >>>>> FYI, kernel build failed on
> >>>>>
> >>>>> tree: git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git multiqueue
> >>>>> head: e9a6aa8a0b7b93195774a9fcf551632bf3abcd97
> >>>>> commit: 67db7275fa4325d46853a4bfcb79acf32e48eca8 [1/3] multiqueue: a hodge podge of things
> >>>>
> >>>> The multiqueue branch is a private branch, it's known broken on many
> >>>> configs at the moment.
> >>>
> >>> OK. Sorry my script just blindly build tests every new/updated branches
> >>> in the tree. It does try to reduce noise level by remembering all warned
> >>> error messages. However if it still presents a problem to the private
> >>> branches, I can either explicitly blacklist "multiqueue", or let the
> >>> script automatically skip commits whose "^Signed-off-by:" is still missing.
> >>
> >> It's not a problem that you're building various branches, as long as you
> >> are expecting a bit of noise :-). In fact I applaud your effort on
> >> expanding the scope of building others branches.
> >>
> >> But it does raise the question of how do we best signal intermediate
> >> branches like this, so they can be avoided until they are ready. The
> >> blacklist doesn't seem like a great choice, since then I have to
> >> remember to remind you when it _is_ ok to build it.
> >>
> >> How about some arbitrary tag? Instead of checking a Signed-off-by, I
> >> could just add a Dont-Auto-Build or something like that. What do you
> >> think?
> >
> > OK, I'll add a grep for "Dont-Auto-Build". If it's found in *any*
> > commit of the branch, the build script will skip the whole branch.
> > Does this make sense to you?
>
> Yeah that'll work for me. A description or tag might be more elegant,
> but this is easier...
Great, let's settle with the in-commit tag "Dont-Auto-Build" for now :)
"git branch --edit-description" could potentially be the most elegant
solution, unfortunately for now there seems no way for me to retrieve
the branch description edited by you..
Another alternative, git tag, has the problem that it goes out of sync
every time you do rebase, which will sure happen a lot for early RFC
patches.
Thanks,
Fengguang
On Thu, 14 Jun 2012 19:31:42 +0800, Fengguang Wu <[email protected]> wrote:
> On Thu, Jun 14, 2012 at 12:45:38PM +0200, Jens Axboe wrote:
> > On 06/14/2012 12:02 PM, Fengguang Wu wrote:
> > > OK, I'll add a grep for "Dont-Auto-Build". If it's found in *any*
> > > commit of the branch, the build script will skip the whole branch.
> > > Does this make sense to you?
> >
> > Yeah that'll work for me. A description or tag might be more elegant,
> > but this is easier...
>
> Great, let's settle with the in-commit tag "Dont-Auto-Build" for now :)
>
> "git branch --edit-description" could potentially be the most elegant
> solution, unfortunately for now there seems no way for me to retrieve
> the branch description edited by you..
>
> Another alternative, git tag, has the problem that it goes out of sync
> every time you do rebase, which will sure happen a lot for early RFC
> patches.
Commit notes or a special namespace for branches (refs/heads/wip/...) would
also be an option.
tom