Make Chromium Mac 10.6 Shared not check for dependencies.

Make Chromium Mac 10.6 Shared not check for dependencies.

It is necessary since it builds all the targets and some targets require test data. Ensure tests_run=noop disable the isolate steps.

R=thakis@chromium.org
BUG=98636
TEST=

[Nico, 2012-04-30 15:26:54]

Does this mean that `ninja all` now runs tests? (see also
https://chromiumcodereview.appspot.com/10069047/ )

[M-A Ruel, 2012-04-30 15:28:54]

On 2012/04/30 15:26:54, Nico wrote:
> Does this mean that `ninja all` now runs tests? (see also
> https://chromiumcodereview.appspot.com/10069047/ )

No, isolate.py --mode's argument defaulted to 'check', which checks that the
test data files are present. But I'm removing that from the default
configuration since it's breaking from everywhere;
https://chromiumcodereview.appspot.com/10261010/

[Nico, 2012-04-30 15:47:33]

On Mon, Apr 30, 2012 at 8:28 AM,  <maruel@chromium.org> wrote:
> On 2012/04/30 15:26:54, Nico wrote:
>>
>> Does this mean that `ninja all` now runs tests? (see also
>> https://chromiumcodereview.appspot.com/10069047/ )
>
>
> No, isolate.py --mode's argument defaulted to 'check', which checks that the
> test data files are present. But I'm removing that from the default
> configuration since it's breaking from everywhere;
> https://chromiumcodereview.appspot.com/10261010/
>
> https://chromiumcodereview.appspot.com/10264010/

So this change here is no longer necessary?

[M-A Ruel, 2012-04-30 16:11:11]

Rewrote the CL.

Now dependent on https://chromiumcodereview.appspot.com/10261010

[Nico, 2012-04-30 16:16:32]

Should 'noop' be the default for this gyp define?

[M-A Ruel, 2012-04-30 16:17:11]

On 2012/04/30 16:16:32, Nico wrote:
> Should 'noop' be the default for this gyp define?

No, I want devs to have the right default when they build FOO_run. This is only
an issue on builders that do not build a specific target.

[Nico, 2012-04-30 16:20:02]

Ah, devs have full checkouts, gotcha.

lgtm

[nsylvain, 2012-04-30 18:17:16]

On 2012/04/30 16:20:02, Nico wrote:
> Ah, devs have full checkouts, gotcha.
Except that they dont? A lot of dev use the override to nto get everything. And
what about internal data? Most devs dont fetch src-internal at all.

> 
> lgtm

[M-A Ruel, 2012-04-30 18:21:00]

On 2012/04/30 18:17:16, nsylvain wrote:
> On 2012/04/30 16:20:02, Nico wrote:
> > Ah, devs have full checkouts, gotcha.
> Except that they dont? A lot of dev use the override to nto get everything.
And
> what about internal data? Most devs dont fetch src-internal at all.

Then they don't run foo_run.

[Nico, 2012-04-30 18:27:41]

On 2012/04/30 18:21:00, Marc-Antoine Ruel wrote:
> On 2012/04/30 18:17:16, nsylvain wrote:
> > On 2012/04/30 16:20:02, Nico wrote:
> > > Ah, devs have full checkouts, gotcha.
> > Except that they dont? A lot of dev use the override to nto get everything.
> And
> > what about internal data? Most devs dont fetch src-internal at all.
> 
> Then they don't run foo_run.

If the _run stuff depends on src-internal, the default should be "noop". Running
'ninja all' should work for developers, without warnings or errors.

[M-A Ruel, 2012-04-30 18:31:47]

On 2012/04/30 18:27:41, Nico wrote:
> On 2012/04/30 18:21:00, Marc-Antoine Ruel wrote:
> > On 2012/04/30 18:17:16, nsylvain wrote:
> > > On 2012/04/30 16:20:02, Nico wrote:
> > > > Ah, devs have full checkouts, gotcha.
> > > Except that they dont? A lot of dev use the override to nto get
everything.
> > And
> > > what about internal data? Most devs dont fetch src-internal at all.
> > 
> > Then they don't run foo_run.
> 
> If the _run stuff depends on src-internal, the default should be "noop".
Running
> 'ninja all' should work for developers, without warnings or errors.

foo_run aren't included in target 'all'.

[nsylvain, 2012-04-30 18:36:11]

On Mon, Apr 30, 2012 at 11:31 AM, <maruel@chromium.org> wrote:

> On 2012/04/30 18:27:41, Nico wrote:
>
>> On 2012/04/30 18:21:00, Marc-Antoine Ruel wrote:
>> > On 2012/04/30 18:17:16, nsylvain wrote:
>> > > On 2012/04/30 16:20:02, Nico wrote:
>> > > > Ah, devs have full checkouts, gotcha.
>> > > Except that they dont? A lot of dev use the override to nto get
>>
> everything.
>
>> > And
>> > > what about internal data? Most devs dont fetch src-internal at all.
>> >
>> > Then they don't run foo_run.
>>
>
>  If the _run stuff depends on src-internal, the default should be "noop".
>>
> Running
>
>> 'ninja all' should work for developers, without warnings or errors.
>>
>
> foo_run aren't included in target 'all'.
>
Then why does not matter at all again?
/me confused.

>
>
https://chromiumcodereview.**appspot.com/10264010/<https://chromiumcodereview...
>

[M-A Ruel, 2012-04-30 18:37:52]

On 2012/04/30 18:36:11, nsylvain wrote:
> On Mon, Apr 30, 2012 at 11:31 AM, <mailto:maruel@chromium.org> wrote:
> > On 2012/04/30 18:27:41, Nico wrote:
> > Running
> >
> >> 'ninja all' should work for developers, without warnings or errors.
> >>
> >
> > foo_run aren't included in target 'all'.
> >
> Then why does not matter at all again?
> /me confused.

Because some builders do not build 'all', they build everything. There's a
difference.

[Nico, 2012-04-30 18:40:03]

As far as I understand, in ninja 'all' is everything, and 'All' is build/all.gyp
(but I'm not sure).

[M-A Ruel, 2012-04-30 18:48:56]

On 2012/04/30 18:40:03, Nico wrote:
> As far as I understand, in ninja 'all' is everything, and 'All' is
build/all.gyp
> (but I'm not sure).

http://martine.github.com/ninja/manual.html doesn't document anything named
"all".

[Nico, 2012-04-30 18:51:07]

Yes, it's written by tools/gyp/pylib/gyp/generator/ninja.py

On Mon, Apr 30, 2012 at 11:48 AM,  <maruel@chromium.org> wrote:
> On 2012/04/30 18:40:03, Nico wrote:
>>
>> As far as I understand, in ninja 'all' is everything, and 'All' is
>
> build/all.gyp
>>
>> (but I'm not sure).
>
>
> http://martine.github.com/ninja/manual.html doesn't document anything named
> "all".
>
> https://chromiumcodereview.appspot.com/10264010/

[M-A Ruel, 2012-04-30 19:32:19]

On 2012/04/30 18:51:07, Nico wrote:
> Yes, it's written by tools/gyp/pylib/gyp/generator/ninja.py

I verified "ninja all" do build base_unittests_run.

It's possible I leave the default to noop then. In any case, it makes sense to
enforce it there as a safeguard.

[M-A Ruel, 2012-05-01 14:25:05]

In any case, I'm abandoning this patch since the noop trick is useless. The
build tool still fails because it fails to find the files in
"isolate_dependency_tracked". So the fix is to move the files not guaranteed to
be present to "isolate_dependency_untracked".

[M-A Ruel, 2012-05-01 14:38:12]

On 2012/05/01 14:25:05, Marc-Antoine Ruel wrote:
> In any case, I'm abandoning this patch since the noop trick is useless. The
> build tool still fails because it fails to find the files in
> "isolate_dependency_tracked". So the fix is to move the files not guaranteed
to
> be present to "isolate_dependency_untracked".

Err, I had a rough morning. While isolate_dependency_untracked is necessary for
potentially missing files, it is not sufficient, it must also use --mode=noop.
So keeping this CL open.