Remove webkit.gyp:pull_in_test_shell

Include test_shell.gypi in webkit.gyp instead of maintaining a distinct
test_shell.gyp.  This allows the removal of webkit.gyp:pull_in_test_shell,
eliminating the circular relationship between webkit.gyp and test_shell.gyp.

The expected semantics of building all of test_shell during a webkit.gyp:All
build is performed are retained.

This change depends on GYP r781 (in Chromium at r38526) for GYP suffix
operator path section treatment.  It also depends on Chromium r38519 and
r38658 for common.gypi treatment of chromium_code.

BUG=33186
TEST=Xcode no longer holds stale project files in memory

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=38664

[Mark Mentovai, 2010-02-09 21:24:12]

For advance feedback.  Dependencies noted, and this will need some try-bot
testing once the dependencies land.

[TVL, 2010-02-09 21:29:01]

so this will break a few bots, we have bots that directly build
test_shell.xcodeproj to cut down on that they build, so landing it means also
doing a master.cfg change and bounce.

[TVL, 2010-02-09 21:29:29]

On 2010/02/09 21:29:01, TVL wrote:
> so this will break a few bots, we have bots that directly build
> test_shell.xcodeproj to cut down on that they build, so landing it means also
> doing a master.cfg change and bounce.

maybe include a dummy test_shell.gyp to redirect to webkit.gyp as a bridge?

[TVL, 2010-02-09 21:34:25]

http://codereview.chromium.org/602006/diff/1/3
File webkit/tools/test_shell/test_shell.gypi (right):

http://codereview.chromium.org/602006/diff/1/3#newcode31
webkit/tools/test_shell/test_shell.gypi:31: # dependencies are relative to
webkit.gyp, which includes this file.
would using DEPTH help avoid any of these comments?

[Mark Mentovai, 2010-02-09 22:15:51]

Darin told me that the bots were changed to just build
webkit.xcodeproj.  Do you know where this is set up?  I'll add a
test_shell.gyp with a dummy target if I need to to avoid needing an
immediate master restart.

Mark

TVL wrote:
> so this will break a few bots, we have bots that directly build
> test_shell.xcodeproj to cut down on that they build, so landing it means
> also
> doing a master.cfg change and bounce.

[darin (slow to review), 2010-02-09 22:25:28]

Nicolas made that change for me.
-Darin

On Tue, Feb 9, 2010 at 2:15 PM, Mark Mentovai <mark@chromium.org> wrote:

> Darin told me that the bots were changed to just build
> webkit.xcodeproj.  Do you know where this is set up?  I'll add a
> test_shell.gyp with a dummy target if I need to to avoid needing an
> immediate master restart.
>
> Mark
>
> TVL wrote:
> > so this will break a few bots, we have bots that directly build
> > test_shell.xcodeproj to cut down on that they build, so landing it means
> > also
> > doing a master.cfg change and bounce.
>

[TVL, 2010-02-09 22:33:37]

fyi and memory masters both seem to have test_shell.xcodeproj

TVL


On Tue, Feb 9, 2010 at 5:25 PM, Darin Fisher <darin@chromium.org> wrote:

> Nicolas made that change for me.
> -Darin
>
>
> On Tue, Feb 9, 2010 at 2:15 PM, Mark Mentovai <mark@chromium.org> wrote:
>
>> Darin told me that the bots were changed to just build
>> webkit.xcodeproj.  Do you know where this is set up?  I'll add a
>> test_shell.gyp with a dummy target if I need to to avoid needing an
>> immediate master restart.
>>
>> Mark
>>
>> TVL wrote:
>> > so this will break a few bots, we have bots that directly build
>> > test_shell.xcodeproj to cut down on that they build, so landing it means
>> > also
>> > doing a master.cfg change and bounce.
>>
>
>

[Mark Mentovai, 2010-02-09 22:38:42]

Should those switch to use webkit.xcodeproj instead now too?  I assume so.

If we can't do that before my stuff lands, I'll provide a stub
test_shell.gyp with the right "none"-type targets depending on the
right bits from test_shell.gypi via webkit.gyp.

Mark

TVL wrote:
> fyi and memory masters both seem to have test_shell.xcodeproj
> TVL
>
> On Tue, Feb 9, 2010 at 5:25 PM, Darin Fisher <darin@chromium.org> wrote:
>>
>> Nicolas made that change for me.
>> -Darin
>>
>> On Tue, Feb 9, 2010 at 2:15 PM, Mark Mentovai <mark@chromium.org> wrote:
>>>
>>> Darin told me that the bots were changed to just build
>>> webkit.xcodeproj.  Do you know where this is set up?  I'll add a
>>> test_shell.gyp with a dummy target if I need to to avoid needing an
>>> immediate master restart.
>>>
>>> Mark
>>>
>>> TVL wrote:
>>> > so this will break a few bots, we have bots that directly build
>>> > test_shell.xcodeproj to cut down on that they build, so landing it
>>> > means
>>> > also
>>> > doing a master.cfg change and bounce.
>>
>
>

[Nicolas Sylvain, 2010-02-09 22:45:15]

On Tue, Feb 9, 2010 at 2:38 PM, Mark Mentovai <mark@chromium.org> wrote:

> Should those switch to use webkit.xcodeproj instead now too?  I assume so.
>
Yes.  I just did not change the valgrind bots yet because they did not need
it, but
there is no harm in doing that.

I will prep a cl.

Nicolas


> If we can't do that before my stuff lands, I'll provide a stub
> test_shell.gyp with the right "none"-type targets depending on the
> right bits from test_shell.gypi via webkit.gyp.
>
> Mark
>
> TVL wrote:
> > fyi and memory masters both seem to have test_shell.xcodeproj
> > TVL
> >
> > On Tue, Feb 9, 2010 at 5:25 PM, Darin Fisher <darin@chromium.org> wrote:
> >>
> >> Nicolas made that change for me.
> >> -Darin
> >>
> >> On Tue, Feb 9, 2010 at 2:15 PM, Mark Mentovai <mark@chromium.org>
> wrote:
> >>>
> >>> Darin told me that the bots were changed to just build
> >>> webkit.xcodeproj.  Do you know where this is set up?  I'll add a
> >>> test_shell.gyp with a dummy target if I need to to avoid needing an
> >>> immediate master restart.
> >>>
> >>> Mark
> >>>
> >>> TVL wrote:
> >>> > so this will break a few bots, we have bots that directly build
> >>> > test_shell.xcodeproj to cut down on that they build, so landing it
> >>> > means
> >>> > also
> >>> > doing a master.cfg change and bounce.
> >>
> >
> >
>

[Mark Mentovai, 2010-02-09 23:37:51]

DEPTH only helps if you need to get back to the top and you're not
sure where you are.  We don't need to get back to the top and we are
sure where we are.

The comments are there because most things in a .gypi are properly
treated as relative to the .gypi, but target descriptions are an
exception: they're relative to the .gyp.

Mark

TVL wrote:
> http://codereview.chromium.org/602006/diff/1/3
> File webkit/tools/test_shell/test_shell.gypi (right):
>
> http://codereview.chromium.org/602006/diff/1/3#newcode31
> webkit/tools/test_shell/test_shell.gypi:31: # dependencies are relative
> to webkit.gyp, which includes this file.
> would using DEPTH help avoid any of these comments?

[Mark Mentovai, 2010-02-10 18:02:04]

New version up, using DEPTH.  I see what you were saying now.  Eliminating all
of those comments is much nicer.

[TVL, 2010-02-10 18:05:54]

On 2010/02/10 18:02:04, Mark Mentovai wrote:
> New version up, using DEPTH.  I see what you were saying now.  Eliminating all
> of those comments is much nicer.

much nicer.  thanks.

lgtm

[Mark Mentovai, 2010-02-10 20:52:18]

Final version up, with a fix for winmm.lib pathname relativization - in this
case, relativization is unwanted.

[TVL, 2010-02-10 21:57:12]

lgtm