Split linux_use_gold_binary into linux_use_bundled_gold and linux_use_bundled_binutils.

Split linux_use_gold_binary into linux_use_bundled_gold and linux_use_bundled_binutils.

linux_use_bundled_binutils adds the binutils to cflags.
linux_use_bundled_gold adds it to the ldflags.  The former
does not work when cross compiling to ARM as the other
binutils do not currently support ARM.  This was causing the
ARM cross compile (for which we do want use the bundled
gold) to break.

I think the word "bundled" works better than "binary" since
it distinguishes from the system gold (which is also a binary).

BUG=363661
TEST=linux_arm_cross_compile trybot

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=264165

[Lei Zhang, 2014-04-15 21:40:31]

https://codereview.chromium.org/239163003/diff/1/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/239163003/diff/1/build/common.gypi#newcode735
build/common.gypi:735: # linux_use_bundled_gold: whether to use the binary
checked into
Since we are here, let's make this a bit clearer. WDYT?

# linux_use_bundled_gold: Whether to use the gold linker binary checked
# into third_party/binutils.
# Force this off via GYP_DEFINES when you are using a custom toolchain
# and need to control -B in ldflags.
# Gold is not used for 32-bit linux builds as it runs out of address 
# space.

https://codereview.chromium.org/239163003/diff/1/build/common.gypi#newcode744
build/common.gypi:744: # linux_use_bundled_binutils: whether to use the binary
binutils
Similar to above, mention one can force this off for -B in cflags.

https://codereview.chromium.org/239163003/diff/1/build/common.gypi#newcode746
build/common.gypi:746: # be used except on x86 and x86-64 (the only two
achitectures
achitectures -> architectures

https://codereview.chromium.org/239163003/diff/1/build/common.gypi#newcode1378
build/common.gypi:1378: ['host_arch=="x64"', {
Why set this when linux_use_bundled_binutils=0?

[Sam Clegg, 2014-04-15 21:54:54]

https://codereview.chromium.org/239163003/diff/1/build/common.gypi
File build/common.gypi (right):

https://codereview.chromium.org/239163003/diff/1/build/common.gypi#newcode735
build/common.gypi:735: # linux_use_bundled_gold: whether to use the binary
checked into
On 2014/04/15 21:40:31, Lei Zhang wrote:
> Since we are here, let's make this a bit clearer. WDYT?
> 
> # linux_use_bundled_gold: Whether to use the gold linker binary checked
> # into third_party/binutils.
> # Force this off via GYP_DEFINES when you are using a custom toolchain
> # and need to control -B in ldflags.
> # Gold is not used for 32-bit linux builds as it runs out of address 
> # space.

Done.

https://codereview.chromium.org/239163003/diff/1/build/common.gypi#newcode744
build/common.gypi:744: # linux_use_bundled_binutils: whether to use the binary
binutils
On 2014/04/15 21:40:31, Lei Zhang wrote:
> Similar to above, mention one can force this off for -B in cflags.

Done.

https://codereview.chromium.org/239163003/diff/1/build/common.gypi#newcode746
build/common.gypi:746: # be used except on x86 and x86-64 (the only two
achitectures
On 2014/04/15 21:40:31, Lei Zhang wrote:
> achitectures -> architectures

Done.

https://codereview.chromium.org/239163003/diff/1/build/common.gypi#newcode1378
build/common.gypi:1378: ['host_arch=="x64"', {
On 2014/04/15 21:40:31, Lei Zhang wrote:
> Why set this when linux_use_bundled_binutils=0?

Its needed when linux_use_bundled_binutils=1 or linux_use_bundled_gold=1... 
what me to add the extra conditional?
No harm in having this set is there?

[Lei Zhang, 2014-04-15 22:42:26]

lgtm

Give mithro a chance to weigh in on this?

[Sam Clegg, 2014-04-15 23:47:32]

On 2014/04/15 22:42:26, Lei Zhang wrote:
> lgtm
> 
> Give mithro a chance to weigh in on this?

Will do. Mithro, can you take a look ASAP, we are trying to resurrect the broken
ARM bots fairly urgently.

[mithro-old, 2014-04-16 00:57:42]

On 2014/04/15 23:47:32, Sam Clegg wrote:
> On 2014/04/15 22:42:26, Lei Zhang wrote:
> > lgtm
> > 
> > Give mithro a chance to weigh in on this?
> 
> Will do. Mithro, can you take a look ASAP, we are trying to resurrect the
broken
> ARM bots fairly urgently.

Hi Sam,

As Lei Zhang is happy with this patch, I'm going to kick of the submit queue.
However, I'm not sure this is the right solution here. 

Where is your ARM builder getting the cross compiler from? I'm assuming that
doesn't actually include gold?

Did you see my comments on https://codereview.chromium.org/196573022/?

> > Not sure how to proceed here.  We could disable use of gold linker on ARM,
but
> > that seems unfortunate.  Would it be possible to split gold from the other
> > binutils?  Or perhaps make the binutil multi-arch?
> 
> Couple of options here;
>  * You could try "linux_use_gold_flags=1 linux_use_gold_binary=0" combination
> here?
>  * Setting binutils_dir to the binutils in your cross compiler?
> 
> We can rebuild binutils to be multiarch enabled, but for it to work properly
> we'd need to apply a bunch of patches from debian. I actually started work on
> doing that but stopped.
> 
> > Also, isn't the linux_use_gold_binary a little miss-named now as it does
more
> > than just use the gold binary now?
> 
> Yes.

Tim 'mithro' Ansell

[Lei Zhang, 2014-04-16 01:22:23]

On 2014/04/16 00:57:42, mithro wrote:
> Where is your ARM builder getting the cross compiler from? I'm assuming that
> doesn't actually include gold?

It downloads the arm sysroot and sets the sysroot gyp variable.

I suspect the arm build uses the bundled gold, which can output to some ARM
targets?

[Lei Zhang, 2014-04-16 01:22:33]

The CQ bit was checked by thestig@chromium.org

[mithro-old, 2014-04-16 01:26:45]

On 2014/04/16 01:22:23, Lei Zhang wrote:
> It downloads the arm sysroot and sets the sysroot gyp variable.
> 
> I suspect the arm build uses the bundled gold, which can output to some ARM
> targets?

Should the sysroot instead have it's own bundled gold/binutils? It seems a
little surprising that everything *but* the linker is coming from the sysroot
(can you build with clang?).

Tim

[Sam Clegg, 2014-04-16 01:31:23]

On 2014/04/16 01:26:45, mithro wrote:
> On 2014/04/16 01:22:23, Lei Zhang wrote:
> > It downloads the arm sysroot and sets the sysroot gyp variable.
> > 
> > I suspect the arm build uses the bundled gold, which can output to some ARM
> > targets?
> 
> Should the sysroot instead have it's own bundled gold/binutils? It seems a
> little surprising that everything *but* the linker is coming from the sysroot
> (can you build with clang?).
> 
> Tim

The ARM built is done using the arm-linux-gnueabihf cross toolchain from
ubuntu/precise.

I'm sure it includes gold, but IIUC the point of using our own gold is that we
can use a more recent version than the one that ships with precice.  Also, I'm
not sure if its possible to select gold over ld.bfd from the command line on a
normal precise toolchain, so we would probably end up with ld.bfd if we didn't
use the bundled gold.

[commit-bot: I haz the power, 2014-04-16 01:41:20]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/sbc@chromium.org/239163003/20001

[commit-bot: I haz the power, 2014-04-16 09:50:27]

Change committed as 264165