gn build changes for v8

gn build changes for v8

mksnapshot wasn't being built correctly for android. cflags/ldflags
need to include m32 when building on x64 but targetting arm.

This matches what the gyp side was doing.

When rolling into chrome I'll need to include
https://codereview.chromium.org/988183003 .

R=jochen@chromium.org
BUG=465456
LOG=Y

[sky, 2015-03-09 21:53:41]

sky@chromium.org changed reviewers:
+ marja@chromium.org

[sky, 2015-03-09 21:53:42]

Not sure if jochen is reviewing while travelling, so picking the two of you.

[jochen (gone - plz use gerrit), 2015-03-10 13:36:19]

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

[jochen (gone - plz use gerrit), 2015-03-10 13:36:20]

the code changes lgtm. But won't that break all the gn builders on the v8
waterfall if it depends on a chrome side change?

+machenbach for rolling question
+dpranke because he looked into that before

[sky, 2015-03-10 15:04:08]

On 2015/03/10 13:36:20, jochen (slow) wrote:
> the code changes lgtm. But won't that break all the gn builders on the v8
> waterfall if it depends on a chrome side change?
> 
> +machenbach for rolling question
> +dpranke because he looked into that before

Indeed it will. I was hoping that I could do the DEPs roll and land
https://codereview.chromium.org/988183003 in one patch. If you think that's a
bad idea, then I can do the following:
. land a patch in the chrome side that adds a variable as to whether the code
path in this patch should be enabled. Say enable_correct_v8_arch in
build/module_args/v8.gni. The patch I land now would set it to false.
. land this patch in the v8 side.
. roll deps.
. Land https://codereview.chromium.org/988183003 and include in it setting
enable_correct_v8_arch to true.
. Land patch in v8 side that removes enable_correct_v8_arch.
. roll deps
. Land patch in chrome side that removes enable_correct_v8_arch.

UGH! This is the safest approach, but the most pain. WDYT?

Of course I need to land https://codereview.chromium.org/984353002/ before any
of this.

[jochen (gone - plz use gerrit), 2015-03-10 15:05:23]

n-sided patches ftw

but in order to not risk burning down all waterfalls, I think we should go for
this

[sky, 2015-03-10 15:54:32]

On 2015/03/10 15:05:23, jochen (slow) wrote:
> n-sided patches ftw
> 
> but in order to not risk burning down all waterfalls, I think we should go for
> this

by this you mean what I outlined, right?

[jochen (gone - plz use gerrit), 2015-03-10 16:03:50]

right

[sky, 2015-03-10 16:21:58]

Updated with new flag landing in chrome side. I can't land this until
https://codereview.chromium.org/984353002/ lands.

[Michael Achenbach, 2015-03-10 19:40:30]

+1 on n-sided patches. A v8 roll revert is possible and more pain if several
other things need to be rolled back in sync.

[Michael Achenbach, 2015-03-10 19:43:24]

You could also manually add v8_android_chromium_gn_dbg as trybot for this CL (CQ
and git cl try don't choose it automatically). We don't have linux dbg and
android release, which possibly lets problems slip through.

[sky, 2015-03-10 20:36:40]

On 2015/03/10 19:43:24, Michael Achenbach wrote:
> You could also manually add v8_android_chromium_gn_dbg as trybot for this CL
(CQ
> and git cl try don't choose it automatically). We don't have linux dbg and
> android release, which possibly lets problems slip through.

Done. The patch is now passing the appropriate gn bots. Michael, as you're in my
time zone could you review?

[sky, 2015-03-10 20:37:15]

On 2015/03/10 20:36:40, sky wrote:
> On 2015/03/10 19:43:24, Michael Achenbach wrote:
> > You could also manually add v8_android_chromium_gn_dbg as trybot for this CL
> (CQ
> > and git cl try don't choose it automatically). We don't have linux dbg and
> > android release, which possibly lets problems slip through.
> 
> Done. The patch is now passing the appropriate gn bots. Michael, as you're in
my
> time zone could you review?

Also, this isn't dependent on https://codereview.chromium.org/984353002/
anymore. It can land as is.

[Dirk Pranke, 2015-03-10 21:20:46]

On 2015/03/10 20:37:15, sky wrote:
> On 2015/03/10 20:36:40, sky wrote:
> > On 2015/03/10 19:43:24, Michael Achenbach wrote:
> > > You could also manually add v8_android_chromium_gn_dbg as trybot for this
CL
> > (CQ
> > > and git cl try don't choose it automatically). We don't have linux dbg and
> > > android release, which possibly lets problems slip through.
> > 
> > Done. The patch is now passing the appropriate gn bots. Michael, as you're
in
> my
> > time zone could you review?
> 
> Also, this isn't dependent on https://codereview.chromium.org/984353002/
> anymore. It can land as is.

sky@ and I discussed this a bit off-CL, and I think there's probably a different
way to
solve this problem that I'm going to attempt this afternoon. Will report back.

[sky, 2015-03-16 16:41:12]

Closing out as Dirk has landed his patch.