Android: make the top level Android makefile smarter.

Android: make the top level Android makefile smarter.

Instead of requiring an environment variable to be set, which has
resulted in issues with things not being rebuilt correctly when the
variable is forgotten, check to see if the gyp-generated Android
makefile has been created, and if so assume we are doing a WebView
build. This file only exists if gyp has been run with
GYP_GENERATORS=android which only happens in the WebView tree.

Also, rename CHROMIUM_WEBVIEW_DIR to CHROMIUM_DIR as the directory
itself is not webview-specific; it's just the Chromium tree root.

BUG=

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=164291

[Torne, 2012-10-24 14:45:24]

Hi Yaron, others,

Are you okay with this method of determining if we're doing a WebView build? I
would like to get away from setting special environment variables at build time
as the Android bots don't want to do that and it's error-prone.

GypAndroid.mk gets generated from the gyp files if you run with
GYP_GENERATORS=android. When snapshotting Chromium into the Android tree this
gets run and GypAndroid.mk is checked in, so it will always exist for WebView.
It shouldn't ever exist for regular Chrome for Android builds either upstream or
downstream.

[joth, 2012-10-24 20:09:58]

lgtm (but would like Yaron to confirm too)

https://codereview.chromium.org/11269007/diff/1/Android.mk
File Android.mk (right):

https://codereview.chromium.org/11269007/diff/1/Android.mk#newcode14
Android.mk:14: # WebView build.
nit: the WebView build is a bit ambiguous. (we should standardize terminology)

"we are doing an in android tree build" or  "as part of the android system
build" ?

[Yaron, 2012-10-24 20:33:50]

lgtm

We shouldn't ever do mmm for non-webview builds so I think this is reasonable.
The only troublesome thing is if you ever once generate artifacts with the
android gyp-generator and then try to switch back, you'd need to do: git clean
-f, right? Because the generated files are in the source tree. Might be useful
to add a log (assuming you can) saying "Found GypAndroid.mk, assuming build
inside Android tree"

[Torne, 2012-10-25 11:12:01]

https://codereview.chromium.org/11269007/diff/1/Android.mk
File Android.mk (right):

https://codereview.chromium.org/11269007/diff/1/Android.mk#newcode14
Android.mk:14: # WebView build.
On 2012/10/24 20:09:58, joth wrote:
> nit: the WebView build is a bit ambiguous. (we should standardize terminology)
> 
> "we are doing an in android tree build" or  "as part of the android system
> build" ?

Changed this to "we are doing the WebView build using the Android build system."

[Torne, 2012-10-25 11:15:50]

On 2012/10/24 20:33:50, Yaron wrote:
> The only troublesome thing is if you ever once generate artifacts with the
> android gyp-generator and then try to switch back, you'd need to do: git clean
> -f, right? Because the generated files are in the source tree.

In normal usage you don't generate these makefiles yourself; they're generated
at the time the code is snapshotted into the Android tree and checked in. So, in
that tree they always exist, and in other places they are unlikely to ever have
been generated.

> Might be useful
> to add a log (assuming you can) saying "Found GypAndroid.mk, assuming build
> inside Android tree"

The only place this could emit a message would be during the Android build,
which is both something they'd probably rather not have, and also in most builds
will immediately be missed amongst the millions of other lines of output. The
case where this would be bad is when you have Chromium checked out as a
subdirectory of an Android tree for a reason other than building WebView, *and*
have generated the WebView makefiles, *and* are building Android after that, and
I don't think there's a good way to warn at that stage.

[joth, 2012-10-25 17:35:09]

lgtm

[Yaron, 2012-10-25 18:05:55]

On 2012/10/25 17:35:09, joth wrote:
> lgtm

lgtm