[mac] Fix cross-compilation for Android from a Mac

[mac] Fix cross-compilation for Android from a Mac
Also, fix markdown so lists appear properly in generated HTML at https://chromium.googlesource.com/v8/v8/+/master/docs/

I had to duplicate some logic for finding Android cross-compile binaries in the Makefile.android since we can't override AR path from gyp.

BUG=
TEST=Run make android_arm.release from a Mac

[chris.a.williams, 2015-10-26 16:30:18]

Description was changed from

==========
[mac] Fix cross-compilation for Android from a Mac
Also, fix markdown so lists appear properly in generated HTML at
https://chromium.googlesource.com/v8/v8/+/master/docs/

BUG=
TEST=Run make android_arm.release from a Mac
==========

to

==========
[mac] Fix cross-compilation for Android from a Mac
Also, fix markdown so lists appear properly in generated HTML at
https://chromium.googlesource.com/v8/v8/+/master/docs/

I had to duplicate some logic for finding Android cross-compile binaries in the
Makefile.android since we can't override AR path from gyp.

BUG=
TEST=Run make android_arm.release from a Mac
==========

[chris.a.williams, 2015-10-26 16:30:18]

chris.a.williams@gmail.com changed reviewers:
+ machenbach@chromium.org

[Michael Achenbach, 2015-10-26 20:00:34]

https://codereview.chromium.org/1413693009/diff/1/Makefile.android
File Makefile.android (right):

https://codereview.chromium.org/1413693009/diff/1/Makefile.android#newcode38
Makefile.android:38: HOST_OS = $(shell uname -s | sed -e
's/Linux/linux/;s/Darwin/darwin/')
A lot of this moved to standalone.gypi a while ago:

https://chromium.googlesource.com/v8/v8/+log/master/Makefile.android

See the last couple of CLs... Are you sure this can't be solved with a pure gyp
patch? We usually build with ninja and this is only affecting make.

[chris.a.williams, 2015-10-26 20:45:09]

On 2015/10/26 20:00:34, Michael Achenbach wrote:
> https://codereview.chromium.org/1413693009/diff/1/Makefile.android
> File Makefile.android (right):
> 
> https://codereview.chromium.org/1413693009/diff/1/Makefile.android#newcode38
> Makefile.android:38: HOST_OS = $(shell uname -s | sed -e
> 's/Linux/linux/;s/Darwin/darwin/')
> A lot of this moved to standalone.gypi a while ago:
> 
> https://chromium.googlesource.com/v8/v8/+log/master/Makefile.android
> 
> See the last couple of CLs... Are you sure this can't be solved with a pure
gyp
> patch? We usually build with ninja and this is only affecting make.

The available docs all pointed to make as the build system:
https://chromium.googlesource.com/v8/v8/+/master/docs/building_with_gyp.md and
https://chromium.googlesource.com/v8/v8/+/master/docs/d8_on_android.md

I assumed that was still the default option, though I do now notice the Ninja
section:
https://chromium.googlesource.com/v8/v8/+/master/docs/building_with_gyp.md#Ninja

From my testing locally, I was not able to override AR from the standalone.gypi
for Make, hence duplicating some of the logic in the Makefile.android.  Looks to
be a make/gyp known issue even baked into the unit tests:
https://chromium.googlesource.com/external/gyp/+/master/test/make_global_sett...

If make isn't the preferred way to build, then the docs could use some updating
to point people at the right way to get the project building for Android/etc
using ninja.  Since make is documented and there are makefiles, and that seems
to be the way that people are trying to build
(https://www.mail-archive.com/v8-users@googlegroups.com/msg09215.html) this may
be of use for people trying to build static v8 for Android from a Mac.

[Michael Achenbach, 2015-10-27 14:47:54]

We still use make locally a lot while the continuous bots use ninja.

https://codereview.chromium.org/1413693009/diff/1/build/standalone.gypi
File build/standalone.gypi (right):

https://codereview.chromium.org/1413693009/diff/1/build/standalone.gypi#newco...
build/standalone.gypi:1161: ['CC.host', '<(host_cc)'],
Not sure if I understand the test in
https://chromium.googlesource.com/external/gyp/+/master/test/make_global_sett...
correctly, but it looks to me that AR could be set here. Did you try adding:
['AR', '<!(/bin/echo -n <(android_toolchain)/ar)'],

[Jakob Kummerow, 2015-10-27 16:53:35]

DBC.

https://codereview.chromium.org/1413693009/diff/1/tools/gyp/v8.gyp
File tools/gyp/v8.gyp (right):

https://codereview.chromium.org/1413693009/diff/1/tools/gyp/v8.gyp#newcode149
tools/gyp/v8.gyp:149: 'standalone_static_library': 1,
I don't think we want non-thin libraries, it's a waste of space.

[chris.a.williams, 2015-10-27 17:00:13]

On 2015/10/27 14:47:54, Michael Achenbach (OOO) wrote:
> We still use make locally a lot while the continuous bots use ninja.
> 
> https://codereview.chromium.org/1413693009/diff/1/build/standalone.gypi
> File build/standalone.gypi (right):
> 
>
https://codereview.chromium.org/1413693009/diff/1/build/standalone.gypi#newco...
> build/standalone.gypi:1161: ['CC.host', '<(host_cc)'],
> Not sure if I understand the test in
>
https://chromium.googlesource.com/external/gyp/+/master/test/make_global_sett...
> correctly, but it looks to me that AR could be set here. Did you try adding:
> ['AR', '<!(/bin/echo -n <(android_toolchain)/ar)'],

You can set AR for ninja, but not for make. That test shows you can change
AR.host anywhere, but can't set AR for make based builds.

[chris.a.williams, 2015-10-27 17:02:05]

On 2015/10/27 16:53:35, Jakob wrote:
> DBC.
> 
> https://codereview.chromium.org/1413693009/diff/1/tools/gyp/v8.gyp
> File tools/gyp/v8.gyp (right):
> 
> https://codereview.chromium.org/1413693009/diff/1/tools/gyp/v8.gyp#newcode149
> tools/gyp/v8.gyp:149: 'standalone_static_library': 1,
> I don't think we want non-thin libraries, it's a waste of space.

If we're generating a static library that we want to ship and use downstream for
other libraries to build against, we want a "fat" static library. Otherwise we
need to ship the thin static library plus all the objects its referencing, or am
I incorrect?

[Michael Achenbach, 2015-10-30 11:42:17]

machenbach@chromium.org changed reviewers:
+ jkummerow@chromium.org, jochen@chromium.org

[Michael Achenbach, 2015-10-30 11:42:17]

Changes to standalone.gypi - ok if they pass CQ.
Changes to Makefile.android - only used for standalone make setup. The code
duplication is sad but ok.
For changes in v8.gyp please get comments from another  v8 developer. +jkummerow
and jochen

[Michael Achenbach, 2015-10-30 11:42:39]

The CQ bit was checked by machenbach@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2015-10-30 11:42:51]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1413693009/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1413693009/1

[commit-bot: I haz the power, 2015-10-30 11:47:21]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-10-30 11:47:22]

Dry run: Try jobs failed on following builders:
  v8_android_arm_compile_rel on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_android_arm_compile_rel/...)

[Jakob Kummerow, 2015-10-30 11:51:08]

On 2015/10/27 17:02:05, chris.a.williams wrote:
> On 2015/10/27 16:53:35, Jakob wrote:
> > DBC.
> > 
> > https://codereview.chromium.org/1413693009/diff/1/tools/gyp/v8.gyp
> > File tools/gyp/v8.gyp (right):
> > 
> >
https://codereview.chromium.org/1413693009/diff/1/tools/gyp/v8.gyp#newcode149
> > tools/gyp/v8.gyp:149: 'standalone_static_library': 1,
> > I don't think we want non-thin libraries, it's a waste of space.
> 
> If we're generating a static library that we want to ship and use downstream
for
> other libraries to build against, we want a "fat" static library. Otherwise we
> need to ship the thin static library plus all the objects its referencing, or
am
> I incorrect?

That's correct; however shipping static libraries is a rare corner case, and
this change as it is makes every V8 build slower and more disk space consuming.
If you insist that you need to ship static libraries (are you sure that's what
you want?), then it should be an optional, off-by-default setting.

[Jakob Kummerow, 2015-10-30 11:57:25]

In light of the tryjob failures:

https://codereview.chromium.org/1413693009/diff/1/build/standalone.gypi
File build/standalone.gypi (right):

https://codereview.chromium.org/1413693009/diff/1/build/standalone.gypi#newco...
build/standalone.gypi:347: 'host_ld': '<!(which ld)',
Yeah, this can't be in an OS-independent section.

https://codereview.chromium.org/1413693009/diff/1/build/standalone.gypi#newco...
build/standalone.gypi:352: 'host_ld': '<!(which ld)',
Also, having the same stuff in both branches of a condition is a clear sign that
it shouldn't be conditional at all.