Fix order of includes in base_nacl.gyp

Fix order of includes in base_nacl.gyp

GYP wasn't generating correct list of dependencies when base.gypi was
included after untrusted.gypi. As result libbase_nacl.a wasn't always
rebuilt when the sources it depends on were changed.

BUG=461422
Committed: https://crrev.com/a446829989946c6231565e9b753c67c56c6feda4
Cr-Commit-Position: refs/heads/master@{#319092}

[Sergey Ulanov, 2015-03-03 22:07:55]

sergeyu@chromium.org changed reviewers:
+ rvargas@chromium.org

[rvargas (doing something else), 2015-03-03 22:31:14]

rvargas@chromium.org changed reviewers:
+ dpranke@chromium.org

[rvargas (doing something else), 2015-03-03 22:31:15]

This looks like a very subtle (fragile?) change. I'll defer to Dirk's wisdom.

[Dirk Pranke, 2015-03-04 00:14:32]

dpranke@chromium.org changed reviewers:
+ bradnelson@chromium.org

[Dirk Pranke, 2015-03-04 00:14:51]

On 2015/03/03 22:31:15, rvargas wrote:
> This looks like a very subtle (fragile?) change. I'll defer to Dirk's wisdom.

I'm passing the buck on this to Bradley :) ...

[bradn, 2015-03-04 00:55:48]

bradnelson@google.com changed reviewers:
+ bradnelson@google.com

[bradn, 2015-03-04 00:55:48]

This is a little worrisome, though not shocking.
Unfortunately the magic used in base_targets likely has order dependencies with
the magic in untrusted.gypi.

A nicer refactoring might be to switch base.gypi to be just a variable
containing the source list, but this is a larger refactoring.

lgtm

[brucedawson, 2015-03-04 01:05:28]

Is common_untrusted.gypi at risk for conflicting with other .gypi includes?
About half of the 24 uses of common_untrusted.gypi have another .gypi file
included afterwards (such as ppapi_sources.gypi from
D:\src\chromium2\src\ppapi\ppapi_nacl.gyp). Just checking...

It would be nice to make the incorrect ordering cause a build failure rather
than a subtle bug, but for a one-off problem it's not worth it.

Thanks for the fix.

[Sergey Ulanov, 2015-03-04 01:06:47]

On 2015/03/04 00:55:48, bradn wrote:
> This is a little worrisome, though not shocking.
> Unfortunately the magic used in base_targets likely has order dependencies
with
> the magic in untrusted.gypi.
> 
> A nicer refactoring might be to switch base.gypi to be just a variable
> containing the source list, but this is a larger refactoring.

I agree. Added TODO.

> 
> lgtm

[Sergey Ulanov, 2015-03-04 01:12:26]

On 2015/03/04 01:05:28, brucedawson wrote:
> Is common_untrusted.gypi at risk for conflicting with other .gypi includes?
> About half of the 24 uses of common_untrusted.gypi have another .gypi file
> included afterwards (such as ppapi_sources.gypi from
> D:\src\chromium2\src\ppapi\ppapi_nacl.gyp). Just checking...

As far as I can tell base_nacl.gypi is the only place where include order
matters. Other NaCl targets include gypi file that defines list of sources,
which is then referenced in the target. But base.gypi works differently - it
uses target_defaults to set list of sources and this conflicts with the
target_default in untrusted.gypi.

> 
> It would be nice to make the incorrect ordering cause a build failure rather
> than a subtle bug, but for a one-off problem it's not worth it.
> 
> Thanks for the fix.

[rvargas (doing something else), 2015-03-04 17:27:08]

rbstmp lgtm

[Sergey Ulanov, 2015-03-04 18:15:50]

The CQ bit was checked by sergeyu@chromium.org

[Sergey Ulanov, 2015-03-04 18:15:50]

The patchset sent to the CQ was uploaded after l-g-t-m from
bradnelson@google.com
Link to the patchset: https://codereview.chromium.org/976863002/#ps20001 (title:
" ")

[commit-bot: I haz the power, 2015-03-04 18:16:57]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/976863002/20001

[commit-bot: I haz the power, 2015-03-04 18:58:24]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-03-04 18:59:05]

Patchset 2 (id:??) landed as
https://crrev.com/a446829989946c6231565e9b753c67c56c6feda4
Cr-Commit-Position: refs/heads/master@{#319092}