Enable ARM/linux cross compile to use clang.

Enable ARM/linux cross compile to use clang.

BUG=395832
NOTRY=true

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=290261

[Nico, 2014-07-22 22:27:44]

https://codereview.chromium.org/408393002/diff/1/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/408393002/diff/1/build/common.gypi#newcode3693
build/common.gypi:3693: # '-no-integrated-as',
Look at what the android build does. I think it uses -fno-integrated-as and adds
a -B flag to point at the right as (?)

[Sam Clegg, 2014-07-22 22:31:28]

https://codereview.chromium.org/408393002/diff/1/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/408393002/diff/1/build/common.gypi#newcode3693
build/common.gypi:3693: # '-no-integrated-as',
On 2014/07/22 22:27:44, Nico (away) wrote:
> Look at what the android build does. I think it uses -fno-integrated-as and
adds
> a -B flag to point at the right as (?)

Yes, but I think android cross compiler is called 'as' where as we want to use
'/usr/bin/arm-linux-gnueabihf-as'.  The sysroot we are using is a native target
sysroot that doesn't have any host tools in it so I don't think we want to point
-B at that.

I have a look at the clang sources and see if we can overide the name of that
'as' that it uses.

[Nico, 2014-07-22 22:34:39]

On 2014/07/22 22:31:28, Sam Clegg wrote:
> https://codereview.chromium.org/408393002/diff/1/build/common.gypi
> File build/common.gypi (right):
> 
> https://codereview.chromium.org/408393002/diff/1/build/common.gypi#newcode3693
> build/common.gypi:3693: # '-no-integrated-as',
> On 2014/07/22 22:27:44, Nico (away) wrote:
> > Look at what the android build does. I think it uses -fno-integrated-as and
> adds
> > a -B flag to point at the right as (?)
> 
> Yes, but I think android cross compiler is called 'as' where as we want to use
> '/usr/bin/arm-linux-gnueabihf-as'.  The sysroot we are using is a native
target
> sysroot that doesn't have any host tools in it so I don't think we want to
point
> -B at that.
> 
> I have a look at the clang sources and see if we can overide the name of that
> 'as' that it uses.

Oh, I see!

Horrible hack if you don't find anything else: Add builds/arm-hack, put a shell
script there that calls /usr/bin/arm-linux-gnueabihf-as , and add -B
build/arm-hack (medium-term, we should make the integrated as work)

[Nico, 2014-07-22 22:35:26]

On 2014/07/22 22:34:39, Nico (away) wrote:
> On 2014/07/22 22:31:28, Sam Clegg wrote:
> > https://codereview.chromium.org/408393002/diff/1/build/common.gypi
> > File build/common.gypi (right):
> > 
> >
https://codereview.chromium.org/408393002/diff/1/build/common.gypi#newcode3693
> > build/common.gypi:3693: # '-no-integrated-as',
> > On 2014/07/22 22:27:44, Nico (away) wrote:
> > > Look at what the android build does. I think it uses -fno-integrated-as
and
> > adds
> > > a -B flag to point at the right as (?)
> > 
> > Yes, but I think android cross compiler is called 'as' where as we want to
use
> > '/usr/bin/arm-linux-gnueabihf-as'.  The sysroot we are using is a native
> target
> > sysroot that doesn't have any host tools in it so I don't think we want to
> point
> > -B at that.
> > 
> > I have a look at the clang sources and see if we can overide the name of
that
> > 'as' that it uses.
> 
> Oh, I see!
> 
> Horrible hack if you don't find anything else: Add builds/arm-hack, put a
shell
> script there that calls /usr/bin/arm-linux-gnueabihf-as , and add -B
> build/arm-hack (medium-term, we should make the integrated as work)

chandlerc says: "clang --target=arm-linux-gnueabihf should look for a
triple-prefixed as binary", so try passing that :-)

[Sam Clegg, 2014-07-22 22:49:16]

On 2014/07/22 22:35:26, Nico (away) wrote:
> On 2014/07/22 22:34:39, Nico (away) wrote:
> > On 2014/07/22 22:31:28, Sam Clegg wrote:
> > > https://codereview.chromium.org/408393002/diff/1/build/common.gypi
> > > File build/common.gypi (right):
> > > 
> > >
> https://codereview.chromium.org/408393002/diff/1/build/common.gypi#newcode3693
> > > build/common.gypi:3693: # '-no-integrated-as',
> > > On 2014/07/22 22:27:44, Nico (away) wrote:
> > > > Look at what the android build does. I think it uses -fno-integrated-as
> and
> > > adds
> > > > a -B flag to point at the right as (?)
> > > 
> > > Yes, but I think android cross compiler is called 'as' where as we want to
> use
> > > '/usr/bin/arm-linux-gnueabihf-as'.  The sysroot we are using is a native
> > target
> > > sysroot that doesn't have any host tools in it so I don't think we want to
> > point
> > > -B at that.
> > > 
> > > I have a look at the clang sources and see if we can overide the name of
> that
> > > 'as' that it uses.
> > 
> > Oh, I see!
> > 
> > Horrible hack if you don't find anything else: Add builds/arm-hack, put a
> shell
> > script there that calls /usr/bin/arm-linux-gnueabihf-as , and add -B
> > build/arm-hack (medium-term, we should make the integrated as work)
> 
> chandlerc says: "clang --target=arm-linux-gnueabihf should look for a
> triple-prefixed as binary", so try passing that :-)

Thanks .. that seems to have worked.. I was using "-target
armv7a-linux-gnueabihf".

[Sam Clegg, 2014-07-23 01:53:44]

'chrome' target now builds, but depends on some NaCl changes.

[Sam Clegg, 2014-08-08 11:55:56]

NaCl changes are now landed.

This change switches the arm build to clang to makes the arm_linux_cross_compile
bot green again.

PTAL

[Nico, 2014-08-08 15:28:44]

lgtm, thanks!

https://codereview.chromium.org/408393002/diff/180001/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/408393002/diff/180001/build/common.gypi#newco...
build/common.gypi:3721: '-Wno-absolute-value',
nit: Normally, this should go into libvpx.gyp instead. But it looks like it
won't be around for long anyway (that bug's already fixed), so ok.

https://codereview.chromium.org/408393002/diff/180001/build/common.gypi#newco...
build/common.gypi:3735: # handle a several of asm files.
Include link to http://crbug.com/124610 (the bug title says "android" but it's
really about arm)

https://codereview.chromium.org/408393002/diff/180001/chrome/test/data/nacl/n...
File chrome/test/data/nacl/nameservice/pm_nameservice_test.cc (right):

https://codereview.chromium.org/408393002/diff/180001/chrome/test/data/nacl/n...
chrome/test/data/nacl/nameservice/pm_nameservice_test.cc:110:
assert(NACL_SRPC_RESULT_OK == rpc_result);
nit: if this can base, `base::ignore_result(rpc_result);` below the assert is
slightly nicer than the attribute imho

[Nico, 2014-08-14 20:09:32]

What's the status here? Can this land?

[Nico, 2014-08-18 15:38:38]

The CQ bit was checked by thakis@chromium.org

[commit-bot: I haz the power, 2014-08-18 15:39:27]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/sbc@chromium.org/408393002/180001

[commit-bot: I haz the power, 2014-08-18 15:41:14]

Committed patchset #10 (180001) as 290261