Android's native_tests.gyp should take WebKit's directory structure into account.

Android's native_tests.gyp should take WebKit's directory structure into account.

This broke the WebKit Chromium Android bot.

BUG=142322
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=151257

[Yaron, 2012-08-11 18:31:10]

You'd need to update build/apk_test.gypi to specify the same variables as
testing/android/native_test.gyp (that's why the trybot is red). Then you also
need to update the webkit targets (DRT, webkit_unit_tests, etc) to also specify
the variables. I think this may be another three-sided commit because we can't
make the antfile use the variable until everyone's providing it.

[Yaron, 2012-08-11 18:42:24]

On 2012/08/11 18:31:10, Yaron wrote:
> You'd need to update build/apk_test.gypi to specify the same variables as
> testing/android/native_test.gyp (that's why the trybot is red). Then you also
> need to update the webkit targets (DRT, webkit_unit_tests, etc) to also
specify
> the variables. I think this may be another three-sided commit because we can't
> make the antfile use the variable until everyone's providing it.

So it would be: update webkit's gyp targets do this same calculation, have that
roll to chromium, and then land this patch with changes mentioned above. When
this rolls back to webkit, we should be good.

[Peter Beverloo, 2012-08-13 13:15:46]

On 2012/08/11 18:31:10, Yaron wrote:
> You'd need to update build/apk_test.gypi to specify the same variables as
> testing/android/native_test.gyp (that's why the trybot is red). Then you also
> need to update the webkit targets (DRT, webkit_unit_tests, etc) to also
specify
> the variables. I think this may be another three-sided commit because we can't
> make the antfile use the variable until everyone's providing it.

Done. I'm trying a new approach here leveraging the fact that properties in ant
are immutable, making this cycle easier.

[Peter Beverloo, 2012-08-13 13:43:46]

This seems to compile fine now. Adding Satish and Torne.

A WebKit-sided build fix is pending this to land:
https://bugs.webkit.org/show_bug.cgi?id=93841

[Torne, 2012-08-13 13:48:27]

LGTM even if it's a bit of a yucky trick.. seems like the easiest way to get
this going again

[Peter Beverloo, 2012-08-13 13:57:15]

On 2012/08/13 13:48:27, Torne wrote:
> LGTM even if it's a bit of a yucky trick.. seems like the easiest way to get
> this going again

Agreed. The builder has rolled green, so I'm landing this manually.

The underlying problem here is that this system works on absolute directories,
while it really should be using relative directories. Furthermore, Chromium
seems to execute "ant" with the out/{Release,Debug}/ directory as the pwd,
whereas WebKit uses the Chromium source's base directory.

[Yaron, 2012-08-13 15:48:53]

lgtm - ya this is reasonable short-term. Thanks.