Add win_shared and linux_shared builder into the default try bot set.

Add win_shared and linux_shared builder into the default try bot set.

Component builds breakages are easy to introduce because you can accidentally add a cross-component dependency without any warning if you're doing a normal static build.

BUG=105750
TEST=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=112034

[Nico, 2011-11-29 19:34:25]

I'm not a good reviewer for this. nsylvain or maruel might be.

[awong, 2011-11-29 19:34:54]

Swapping reviewers to nsylvain.

[Dirk Pranke, 2011-11-29 21:29:52]

change lgtm but nsylvain or maruel should approve this to make sure we don't
swamp the try bots.

[M-A Ruel, 2011-11-29 21:31:14]

On 2011/11/29 21:29:52, Dirk Pranke wrote:
> change lgtm but nsylvain or maruel should approve this to make sure we don't
> swamp the try bots.

I still think only one of win_shared or win should be kept. Windows slaves don't
scale as well as linux slaves.

lgtm

[awong, 2011-11-29 21:34:20]

On 2011/11/29 21:31:14, Marc-Antoine Ruel wrote:
> On 2011/11/29 21:29:52, Dirk Pranke wrote:
> > change lgtm but nsylvain or maruel should approve this to make sure we don't
> > swamp the try bots.
> 
> I still think only one of win_shared or win should be kept. Windows slaves
don't
> scale as well as linux slaves.
> 
> lgtm

win_shared doesn't run unittests.  If the load proves too high, I'd be happy
even if we could only run linux_shared.  That'd still give us coverage of a
great number of cases where we have cross-module dependencies.

[awong, 2011-11-29 21:35:18]

FYI, committing.  Can revert if it kills our try cycle times.

On 2011/11/29 21:34:20, awong wrote:
> On 2011/11/29 21:31:14, Marc-Antoine Ruel wrote:
> > On 2011/11/29 21:29:52, Dirk Pranke wrote:
> > > change lgtm but nsylvain or maruel should approve this to make sure we
don't
> > > swamp the try bots.
> > 
> > I still think only one of win_shared or win should be kept. Windows slaves
> don't
> > scale as well as linux slaves.
> > 
> > lgtm
> 
> win_shared doesn't run unittests.  If the load proves too high, I'd be happy
> even if we could only run linux_shared.  That'd still give us coverage of a
> great number of cases where we have cross-module dependencies.

[nsylvain, 2011-11-30 22:27:02]

I'm not sure why this would ever work.  We are already hitting the capacity with
our current usage, and end up having pending builds once in a while at the end
of some afternoons.

Right now (still beginning of the afternoon), we have 27 pending jobs on
windows, and 20 pending jobs on linux. This is not really acceptable, and this
will get much worse in the next 2 hours if this day is like any other day.

Nicolas
On 2011/11/29 21:35:18, awong wrote:
> FYI, committing.  Can revert if it kills our try cycle times.
> 
> On 2011/11/29 21:34:20, awong wrote:
> > On 2011/11/29 21:31:14, Marc-Antoine Ruel wrote:
> > > On 2011/11/29 21:29:52, Dirk Pranke wrote:
> > > > change lgtm but nsylvain or maruel should approve this to make sure we
> don't
> > > > swamp the try bots.
> > > 
> > > I still think only one of win_shared or win should be kept. Windows slaves
> > don't
> > > scale as well as linux slaves.
> > > 
> > > lgtm
> > 
> > win_shared doesn't run unittests.  If the load proves too high, I'd be happy
> > even if we could only run linux_shared.  That'd still give us coverage of a
> > great number of cases where we have cross-module dependencies.

[awong, 2011-11-30 22:28:34]

okay...so do you want me to revert?

On Wed, Nov 30, 2011 at 2:27 PM, <nsylvain@google.com> wrote:

> I'm not sure why this would ever work.  We are already hitting the
> capacity with
> our current usage, and end up having pending builds once in a while at the
> end
> of some afternoons.
>
> Right now (still beginning of the afternoon), we have 27 pending jobs on
> windows, and 20 pending jobs on linux. This is not really acceptable, and
> this
> will get much worse in the next 2 hours if this day is like any other day.
>
> Nicolas
>
> On 2011/11/29 21:35:18, awong wrote:
>
>> FYI, committing.  Can revert if it kills our try cycle times.
>>
>
>  On 2011/11/29 21:34:20, awong wrote:
>> > On 2011/11/29 21:31:14, Marc-Antoine Ruel wrote:
>> > > On 2011/11/29 21:29:52, Dirk Pranke wrote:
>> > > > change lgtm but nsylvain or maruel should approve this to make sure
>> we
>> don't
>> > > > swamp the try bots.
>> > >
>> > > I still think only one of win_shared or win should be kept. Windows
>> slaves
>> > don't
>> > > scale as well as linux slaves.
>> > >
>> > > lgtm
>> >
>> > win_shared doesn't run unittests.  If the load proves too high, I'd be
>> happy
>> > even if we could only run linux_shared.  That'd still give us coverage
>> of a
>> > great number of cases where we have cross-module dependencies.
>>
>
>
>
>
http://codereview.chromium.**org/8726043/<http://codereview.chromium.org/8726...
>

[awong, 2011-11-30 22:38:22]

Reverting with drover:
  
 http://codereview.chromium.org/8762005

sad.

-Albert

On 2011/11/30 22:28:34, awong wrote:
> okay...so do you want me to revert?
> 
> On Wed, Nov 30, 2011 at 2:27 PM, <mailto:nsylvain@google.com> wrote:
> 
> > I'm not sure why this would ever work.  We are already hitting the
> > capacity with
> > our current usage, and end up having pending builds once in a while at the
> > end
> > of some afternoons.
> >
> > Right now (still beginning of the afternoon), we have 27 pending jobs on
> > windows, and 20 pending jobs on linux. This is not really acceptable, and
> > this
> > will get much worse in the next 2 hours if this day is like any other day.
> >
> > Nicolas
> >
> > On 2011/11/29 21:35:18, awong wrote:
> >
> >> FYI, committing.  Can revert if it kills our try cycle times.
> >>
> >
> >  On 2011/11/29 21:34:20, awong wrote:
> >> > On 2011/11/29 21:31:14, Marc-Antoine Ruel wrote:
> >> > > On 2011/11/29 21:29:52, Dirk Pranke wrote:
> >> > > > change lgtm but nsylvain or maruel should approve this to make sure
> >> we
> >> don't
> >> > > > swamp the try bots.
> >> > >
> >> > > I still think only one of win_shared or win should be kept. Windows
> >> slaves
> >> > don't
> >> > > scale as well as linux slaves.
> >> > >
> >> > > lgtm
> >> >
> >> > win_shared doesn't run unittests.  If the load proves too high, I'd be
> >> happy
> >> > even if we could only run linux_shared.  That'd still give us coverage
> >> of a
> >> > great number of cases where we have cross-module dependencies.
> >>
> >
> >
> >
> >
>
http://codereview.chromium.**org/8726043/%3Chttp://codereview.chromium.org/87...>
> >