A working version of dart on Android.

A working version of dart_no_snapshot on Android.

BUG=
R=gram@google.com, sgjesse@google.com

Committed: https://code.google.com/p/dart/source/detail?r=23335

[Bill Hesse, 2013-05-28 16:43:41]

The code to generate the snapshot on Android has been removed - dartlib-withcore
now loads all the library source code from the original files, so all the files
would need to be copied.  We think it is best to generate the snapshot on the
host instead.

I just noticed, that dart_no_snapshot will of course not work - I'll verify
that.

[zra, 2013-05-28 16:56:16]

DBC

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi
File runtime/bin/bin.gypi (right):

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi#newco...
runtime/bin/bin.gypi:537: 'toolsets':['target'],
To make running tests easier, we'll likely also be building run_vm_tests for the
host architecture, which requires its own build target. I have that going under
this CL: https://codereview.chromium.org/16060004/. After that, it might make
sense to make a variable for the Android link settings rather than copying it a
third time, if possible.

[gram, 2013-05-28 17:04:19]

lgtm with minor cleanup

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi
File runtime/bin/bin.gypi (right):

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi#newco...
runtime/bin/bin.gypi:429: # TODO(gram): We need to support debug vs release
modes.
I think this comment can just be removed - it refers to the
skia_libs_location_android setting which is commented out

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi#newco...
runtime/bin/bin.gypi:509: # libraries used by the Skia build for now.
Ditto

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi#newco...
runtime/bin/bin.gypi:593: # TODO(gram): We need to support debug vs release
modes.
Ditto

[Bill Hesse, 2013-05-29 09:12:43]

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi
File runtime/bin/bin.gypi (right):

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi#newco...
runtime/bin/bin.gypi:429: # TODO(gram): We need to support debug vs release
modes.
On 2013/05/28 17:04:19, gram wrote:
> I think this comment can just be removed - it refers to the
> skia_libs_location_android setting which is commented out

Done.

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi#newco...
runtime/bin/bin.gypi:509: # libraries used by the Skia build for now.
On 2013/05/28 17:04:19, gram wrote:
> Ditto

Done.

https://codereview.chromium.org/15706008/diff/8001/runtime/bin/bin.gypi#newco...
runtime/bin/bin.gypi:537: 'toolsets':['target'],
On 2013/05/28 16:56:16, zra wrote:
> To make running tests easier, we'll likely also be building run_vm_tests for
the
> host architecture, which requires its own build target. I have that going
under
> this CL: https://codereview.chromium.org/16060004/. After that, it might make
> sense to make a variable for the Android link settings rather than copying it
a
> third time, if possible.

I wasn't sure if the entire 'link_settings' section can be a variable, or only
flat lists, so I didn't make a variable yet.
Merged with your CL by putting the android linking only on the target target,
not the host target.

[Søren Gjesse, 2013-05-29 09:20:42]

lgtm

https://codereview.chromium.org/15706008/diff/8001/runtime/tools/create_snaps...
File runtime/tools/create_snapshot_bin.py (left):

https://codereview.chromium.org/15706008/diff/8001/runtime/tools/create_snaps...
runtime/tools/create_snapshot_bin.py:62: def RunAdb(device, command):
Nice!

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi
File runtime/bin/bin.gypi (right):

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi#newc...
runtime/bin/bin.gypi:345: 'toolsets':['host','target'],
In most of the other places the order is reversed 'target','host'.

[Bill Hesse, 2013-05-29 09:43:19]

Committed patchset #5 manually as r23335 (presubmit successful).

[Ivan Posva, 2013-05-30 20:34:54]

There is still some cleanup pending here.

-Ivan

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi
File runtime/bin/bin.gypi (right):

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi#newc...
runtime/bin/bin.gypi:202: 'io_impl_sources_no_nss.gypi',
Where is 'io_impl_sources_no_nss.gypi'?

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi#newc...
runtime/bin/bin.gypi:345: 'toolsets':['host','target'],
I would like to understand why the generate_resources_cc_file had to get a
toolsets property. We were successfully cross-building before this was added.

What is special about the generate_resources_cc_file that needs to run for the
host specifically and why did it not fail for us when cross-building for ARM or
MIPS?

Even if the 'host' toolset is necessary for some reason, then the 'target' is
unneeded because it is only being used with the #host tag.

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi#newc...
runtime/bin/bin.gypi:374: 'toolsets':['target'],
This seems rather unnecessary, 'target' is the default toolset. Why was this
needed?

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi#newc...
runtime/bin/bin.gypi:380: 'generate_resources_cc_file#host',
I am surprised that the #host was necessary here.

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi#newc...
runtime/bin/bin.gypi:444: 'toolsets':['target'],
ditto

[Bill Hesse, 2013-05-31 15:40:43]

The requested changes are in

https://codereview.chromium.org/16254005/

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi
File runtime/bin/bin.gypi (right):

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi#newc...
runtime/bin/bin.gypi:202: 'io_impl_sources_no_nss.gypi',
On 2013/05/30 20:34:54, Ivan Posva wrote:
> Where is 'io_impl_sources_no_nss.gypi'?

This was accidentally omitted from the CL.  It is in CL
https://codereview.chromium.org/16122004/.

It should be possible to merge io_impl_sources.gypi and
io_impl_sources_no_nss.gypi, since they are almost identical.
I will include this in the cleanup CL.

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi#newc...
runtime/bin/bin.gypi:345: 'toolsets':['host','target'],
On 2013/05/30 20:34:54, Ivan Posva wrote:
> I would like to understand why the generate_resources_cc_file had to get a
> toolsets property. We were successfully cross-building before this was added.
> 
> What is special about the generate_resources_cc_file that needs to run for the
> host specifically and why did it not fail for us when cross-building for ARM
or
> MIPS?
> 
> Even if the 'host' toolset is necessary for some reason, then the 'target' is
> unneeded because it is only being used with the #host tag.

Done.

https://codereview.chromium.org/15706008/diff/16001/runtime/bin/bin.gypi#newc...
runtime/bin/bin.gypi:380: 'generate_resources_cc_file#host',
On 2013/05/30 20:34:54, Ivan Posva wrote:
> I am surprised that the #host was necessary here.

Done.