Revert the part of r90660 that enables building ppapi examples on win.

Revert the part of r90660 that enables building ppapi examples on win.

That change broke the Win(clobber) and WinShared(dbg) builds in the compile
phase.  Examples:
http://build.chromium.org/p/chromium/builders/Win/builds/5309
http://build.chromium.org/p/chromium/builders/Win%20Builder%20%28dbg%29%28shared%29/builds/9605

BUG=54005
TEST=not so much, since the trybots don't build all.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=90712

[Ryan Sleevi, 2011-06-28 04:01:04]

LGTM.

[Ryan Sleevi, 2011-06-28 04:03:02]

As mentioned on IRC, the following is the sequence of dependencies that are
causing trouble:

The 'all' target in src/build/all.gyp
http://codesearch.google.com/codesearch#OAMlx_jo-ck/src/build/all.gyp&l=25
The 'pull_in_all' target of src/chrome/chrome.gyp
http://codesearch.google.com/codesearch#OAMlx_jo-ck/src/chrome/chrome.gyp&l=1356

Both targets include src/ppapi/ppapi_internal.gyp:*
src/ppapi/ppapi_internal.gyp includes ppapi_tests.gypi
http://codesearch.google.com/codesearch#OAMlx_jo-ck/src/ppapi/ppapi_internal....

ppapi_tests.gypi is marked in this CL.

As a result, both targets end up being marked as direct dependents of
'ppapi_example_skeleton', which causes the 'direct_dependent_settings'
dictionary to be merged in.

One solution would be to move ppapi_example_skeleton into it's own gyp file
(ppapi_really_internal.gyp ) and then add that as a dependency for the examples
in ppapi_tests.gypi. The :* gyp operator won't pull in the really_internal.gyp,
only the targets in that ppapi_tests.gypi, so you should be fine.

A hackier solution, if it turns out it needs to be in ppapi_tests.gypi (perhaps
there's something google-internal that depends on this .gypi in it's exact form)
would be to modify ppapi_tests.gypi to have a 'ppapi_examples' target which just
included the examples (not the _skeleton), then modify all.gyp/chrome.gyp to
explicitly pull in targets from ppapi_internal.gyp. This isn't really a
maintainable solution, but it should at least provide an alternative for
discussion.

[darin (slow to review), 2011-06-28 05:15:36]

Thank you for the help, and sorry for the bustage!
-Darin

On Mon, Jun 27, 2011 at 9:03 PM, <rsleevi@chromium.org> wrote:

>
> As mentioned on IRC, the following is the sequence of dependencies that are
> causing trouble:
>
> The 'all' target in src/build/all.gyp
> http://codesearch.google.com/**codesearch#OAMlx_jo-ck/src/**
>
build/all.gyp&l=25<http://codesearch.google.com/codesearch#OAMlx_jo-ck/src/build/all.gyp&l=25>
> The 'pull_in_all' target of src/chrome/chrome.gyp
> http://codesearch.google.com/**codesearch#OAMlx_jo-ck/src/**
>
chrome/chrome.gyp&l=1356<http://codesearch.google.com/codesearch#OAMlx_jo-ck/src/chrome/chrome.gyp&l=1356>
>
> Both targets include src/ppapi/ppapi_internal.gyp:*
> src/ppapi/ppapi_internal.gyp includes ppapi_tests.gypi
> http://codesearch.google.com/**codesearch#OAMlx_jo-ck/src/**
>
ppapi/ppapi_internal.gyp&l=29<http://codesearch.google.com/codesearch#OAMlx_jo-ck/src/ppapi/ppapi_internal.gyp&l=29>
>
> ppapi_tests.gypi is marked in this CL.
>
> As a result, both targets end up being marked as direct dependents of
> 'ppapi_example_skeleton', which causes the 'direct_dependent_settings'
> dictionary to be merged in.
>
> One solution would be to move ppapi_example_skeleton into it's own gyp file
> (ppapi_really_internal.gyp ) and then add that as a dependency for the
> examples
> in ppapi_tests.gypi. The :* gyp operator won't pull in the
> really_internal.gyp,
> only the targets in that ppapi_tests.gypi, so you should be fine.
>
> A hackier solution, if it turns out it needs to be in ppapi_tests.gypi
> (perhaps
> there's something google-internal that depends on this .gypi in it's exact
> form)
> would be to modify ppapi_tests.gypi to have a 'ppapi_examples' target which
> just
> included the examples (not the _skeleton), then modify all.gyp/chrome.gyp
> to
> explicitly pull in targets from ppapi_internal.gyp. This isn't really a
> maintainable solution, but it should at least provide an alternative for
> discussion.
>
>
>
http://codereview.chromium.**org/7230039/<http://codereview.chromium.org/7230...
>

[Ryan Sleevi, 2011-07-11 23:30:01]

On 2011/06/28 05:15:36, darin wrote:
> Thank you for the help, and sorry for the bustage!
> -Darin

mmentovai pointed out an easier, GYPy way to deal with this:

Add 'supress_wildcard': 1 to the ppapi_example_skeleton target to exclude it
from being included by the ppapi_internal.gyp:* include. This will allow the
stubs and examples to be built, while preventing the skeleton dependencies from
being injected into the 'all' target.

[Ami GONE FROM CHROMIUM, 2011-07-12 00:06:35]

On 2011/07/11 23:30:01, Ryan Sleevi wrote:
> On 2011/06/28 05:15:36, darin wrote:
> > Thank you for the help, and sorry for the bustage!
> > -Darin
> 
> mmentovai pointed out an easier, GYPy way to deal with this:
> 
> Add 'supress_wildcard': 1 to the ppapi_example_skeleton target to exclude it
> from being included by the ppapi_internal.gyp:* include. This will allow the
> stubs and examples to be built, while preventing the skeleton dependencies
from
> being injected into the 'all' target.

Thanks Ryan.  Sent http://codereview.chromium.org/7341001/