Add support for iOS to libjpeg.gyp.

Add support for iOS to libjpeg.gyp.

iOS will start pulling libjpeg_turbo in
https://codereview.chromium.org/376573003/ as part of enabling initial Mojo
support on iOS. To do so, iOS needs the same dependency from libjpeg_turbo to
yasm as Mac does; otherwise, GYP chokes in processing libjpeg.gyp due to
yasm_path being undefined.

R=noel@chromium.org, rmcilroy@chromium.org

Committed: 288018

[blundell, 2014-08-06 07:28:56]

Hi Ross,

It's not clear to me who the best reviewer would be for this simple change; I
picked you as you made some changes to this file recently :). Let me know if
there's someone else you think would be better.

Thanks,

Colin

[rmcilroy, 2014-08-06 09:22:09]

Adding Noel who does most of the reviews on libjpeg_turbo.

Are you sure this is required for iOS? I think yasm is only used for x86/x86_64,
so shouldn't be used by an iOS build.  I had a change in-review (which is
currently on hold for other reasons) to cherry-pick arm64 support to
libjpeg_turbo and as part of that I modified the "assemble" target to have the
condition 'target_arch=="ia32" or target_arch=="x64"' instead of
'target_arch!="arm"', would this work?  The change in-question is here:
https://codereview.chromium.org/434123003/

Do you use libjpeg_turbo on iOS?  I was under the impression that this wasn't
used.

[blundell, 2014-08-06 09:30:09]

On 2014/08/06 09:22:09, rmcilroy wrote:
> Adding Noel who does most of the reviews on libjpeg_turbo.
> 
> Are you sure this is required for iOS? I think yasm is only used for
x86/x86_64,
> so shouldn't be used by an iOS build.  I had a change in-review (which is
> currently on hold for other reasons) to cherry-pick arm64 support to
> libjpeg_turbo and as part of that I modified the "assemble" target to have the
> condition 'target_arch=="ia32" or target_arch=="x64"' instead of
> 'target_arch!="arm"', would this work?  The change in-question is here:
> https://codereview.chromium.org/434123003/
> 
> Do you use libjpeg_turbo on iOS?  I was under the impression that this wasn't
> used.

It's a bit of a complicated story. It won't be used, no. The issue is that
processing mojo.gyp requires v8.gyp to be processed (*), which results in a
chain of dependencies that iOS was previously None-ing out needing to be present
in order for gyp to avoid choking, one of them being libjpeg_turbo.

(*) When processing any target in a gypfile, gyp processes *all* the targets in
that gypfile, so even though iOS won't use Mojo targets that depend on v8,
introducing a dependence on mojo.gyp on iOS requires that v8.gyp be present. The
mojo team was unwilling to restructure mojo.gyp to have only the targets that
iOS needs be defined on iOS.

[rmcilroy, 2014-08-06 09:58:01]

On 2014/08/06 09:30:09, blundell wrote:
> On 2014/08/06 09:22:09, rmcilroy wrote:
> > Adding Noel who does most of the reviews on libjpeg_turbo.
> > 
> > Are you sure this is required for iOS? I think yasm is only used for
> x86/x86_64,
> > so shouldn't be used by an iOS build.  I had a change in-review (which is
> > currently on hold for other reasons) to cherry-pick arm64 support to
> > libjpeg_turbo and as part of that I modified the "assemble" target to have
the
> > condition 'target_arch=="ia32" or target_arch=="x64"' instead of
> > 'target_arch!="arm"', would this work?  The change in-question is here:
> > https://codereview.chromium.org/434123003/
> > 
> > Do you use libjpeg_turbo on iOS?  I was under the impression that this
wasn't
> > used.
> 
> It's a bit of a complicated story. It won't be used, no. The issue is that
> processing mojo.gyp requires v8.gyp to be processed (*), which results in a
> chain of dependencies that iOS was previously None-ing out needing to be
present
> in order for gyp to avoid choking, one of them being libjpeg_turbo.
> 
> (*) When processing any target in a gypfile, gyp processes *all* the targets
in
> that gypfile, so even though iOS won't use Mojo targets that depend on v8,
> introducing a dependence on mojo.gyp on iOS requires that v8.gyp be present.
The
> mojo team was unwilling to restructure mojo.gyp to have only the targets that
> iOS needs be defined on iOS.

Ok this makes sense (although it's a bit unfortunate that the mojo team won't
restructure their gyp file to accommodate).  Did the target_arch change I
suggested work, or do you actually build iOS targets with x86 target_arch (the
emulator maybe?)?

[blundell, 2014-08-06 11:00:18]

On 2014/08/06 09:58:01, rmcilroy wrote:
> On 2014/08/06 09:30:09, blundell wrote:
> > On 2014/08/06 09:22:09, rmcilroy wrote:
> > > Adding Noel who does most of the reviews on libjpeg_turbo.
> > > 
> > > Are you sure this is required for iOS? I think yasm is only used for
> > x86/x86_64,
> > > so shouldn't be used by an iOS build.  I had a change in-review (which is
> > > currently on hold for other reasons) to cherry-pick arm64 support to
> > > libjpeg_turbo and as part of that I modified the "assemble" target to have
> the
> > > condition 'target_arch=="ia32" or target_arch=="x64"' instead of
> > > 'target_arch!="arm"', would this work?  The change in-question is here:
> > > https://codereview.chromium.org/434123003/
> > > 
> > > Do you use libjpeg_turbo on iOS?  I was under the impression that this
> wasn't
> > > used.
> > 
> > It's a bit of a complicated story. It won't be used, no. The issue is that
> > processing mojo.gyp requires v8.gyp to be processed (*), which results in a
> > chain of dependencies that iOS was previously None-ing out needing to be
> present
> > in order for gyp to avoid choking, one of them being libjpeg_turbo.
> > 
> > (*) When processing any target in a gypfile, gyp processes *all* the targets
> in
> > that gypfile, so even though iOS won't use Mojo targets that depend on v8,
> > introducing a dependence on mojo.gyp on iOS requires that v8.gyp be present.
> The
> > mojo team was unwilling to restructure mojo.gyp to have only the targets
that
> > iOS needs be defined on iOS.
> 
> Ok this makes sense (although it's a bit unfortunate that the mojo team won't
> restructure their gyp file to accommodate).  Did the target_arch change I
> suggested work, or do you actually build iOS targets with x86 target_arch (the
> emulator maybe?)?

Yep, the simulator is x86.

[rmcilroy, 2014-08-06 11:28:59]

On 2014/08/06 11:00:18, blundell wrote:
> On 2014/08/06 09:58:01, rmcilroy wrote:
> > On 2014/08/06 09:30:09, blundell wrote:
> > > On 2014/08/06 09:22:09, rmcilroy wrote:
> > > > Adding Noel who does most of the reviews on libjpeg_turbo.
> > > > 
> > > > Are you sure this is required for iOS? I think yasm is only used for
> > > x86/x86_64,
> > > > so shouldn't be used by an iOS build.  I had a change in-review (which
is
> > > > currently on hold for other reasons) to cherry-pick arm64 support to
> > > > libjpeg_turbo and as part of that I modified the "assemble" target to
have
> > the
> > > > condition 'target_arch=="ia32" or target_arch=="x64"' instead of
> > > > 'target_arch!="arm"', would this work?  The change in-question is here:
> > > > https://codereview.chromium.org/434123003/
> > > > 
> > > > Do you use libjpeg_turbo on iOS?  I was under the impression that this
> > wasn't
> > > > used.
> > > 
> > > It's a bit of a complicated story. It won't be used, no. The issue is that
> > > processing mojo.gyp requires v8.gyp to be processed (*), which results in
a
> > > chain of dependencies that iOS was previously None-ing out needing to be
> > present
> > > in order for gyp to avoid choking, one of them being libjpeg_turbo.
> > > 
> > > (*) When processing any target in a gypfile, gyp processes *all* the
targets
> > in
> > > that gypfile, so even though iOS won't use Mojo targets that depend on v8,
> > > introducing a dependence on mojo.gyp on iOS requires that v8.gyp be
present.
> > The
> > > mojo team was unwilling to restructure mojo.gyp to have only the targets
> that
> > > iOS needs be defined on iOS.
> > 
> > Ok this makes sense (although it's a bit unfortunate that the mojo team
won't
> > restructure their gyp file to accommodate).  Did the target_arch change I
> > suggested work, or do you actually build iOS targets with x86 target_arch
(the
> > emulator maybe?)?
> 
> Yep, the simulator is x86.

lgtm (though I'm not sure my lgtm is enough or if you need Noel too).

[blundell, 2014-08-06 11:47:17]

On 2014/08/06 11:28:59, rmcilroy wrote:
> On 2014/08/06 11:00:18, blundell wrote:
> > On 2014/08/06 09:58:01, rmcilroy wrote:
> > > On 2014/08/06 09:30:09, blundell wrote:
> > > > On 2014/08/06 09:22:09, rmcilroy wrote:
> > > > > Adding Noel who does most of the reviews on libjpeg_turbo.
> > > > > 
> > > > > Are you sure this is required for iOS? I think yasm is only used for
> > > > x86/x86_64,
> > > > > so shouldn't be used by an iOS build.  I had a change in-review (which
> is
> > > > > currently on hold for other reasons) to cherry-pick arm64 support to
> > > > > libjpeg_turbo and as part of that I modified the "assemble" target to
> have
> > > the
> > > > > condition 'target_arch=="ia32" or target_arch=="x64"' instead of
> > > > > 'target_arch!="arm"', would this work?  The change in-question is
here:
> > > > > https://codereview.chromium.org/434123003/
> > > > > 
> > > > > Do you use libjpeg_turbo on iOS?  I was under the impression that this
> > > wasn't
> > > > > used.
> > > > 
> > > > It's a bit of a complicated story. It won't be used, no. The issue is
that
> > > > processing mojo.gyp requires v8.gyp to be processed (*), which results
in
> a
> > > > chain of dependencies that iOS was previously None-ing out needing to be
> > > present
> > > > in order for gyp to avoid choking, one of them being libjpeg_turbo.
> > > > 
> > > > (*) When processing any target in a gypfile, gyp processes *all* the
> targets
> > > in
> > > > that gypfile, so even though iOS won't use Mojo targets that depend on
v8,
> > > > introducing a dependence on mojo.gyp on iOS requires that v8.gyp be
> present.
> > > The
> > > > mojo team was unwilling to restructure mojo.gyp to have only the targets
> > that
> > > > iOS needs be defined on iOS.
> > > 
> > > Ok this makes sense (although it's a bit unfortunate that the mojo team
> won't
> > > restructure their gyp file to accommodate).  Did the target_arch change I
> > > suggested work, or do you actually build iOS targets with x86 target_arch
> (the
> > > emulator maybe?)?
> > 
> > Yep, the simulator is x86.
> 
> lgtm (though I'm not sure my lgtm is enough or if you need Noel too).

Noel, I'll wait for your review. Thanks!

[Noel Gordon, 2014-08-06 15:39:56]

lgtm - appreciate the details given above, thanks for explaining.

[blundell, 2014-08-07 06:50:00]

On 2014/08/06 15:39:56, Noel Gordon wrote:
> lgtm - appreciate the details given above, thanks for explaining.

Thanks for the reviews. I'm having problems committing:

$ git svn init http://src.chromium.org/svn/trunk/deps/third_party/libjpeg_turbo/
$ git svn fetch
<snip>
$ git cl dcommit
ERROR from SVN:
RA layer request failed: MKACTIVITY request on
'/svn/!svn/act/8e007280-27a1-433c-9b60-0e90a416df5a' failed: 500 Internal Server
Error
ERROR: Not all changes have been committed into SVN.

Any insights?

Thanks,

Colin

[Noel Gordon, 2014-08-07 07:58:24]

Check you have done all steps listed here ...

 
http://www.chromium.org/developers/how-tos/get-the-code#TOC-Working-with-repo...

[Noel Gordon, 2014-08-07 08:06:21]

And if you have done them all, and your git cl dcommit still fails, ping
iannucci

[blundell, 2014-08-07 09:51:57]

Committed patchset #1 manually as 288018 (presubmit successful).

[blundell, 2014-08-07 13:33:14]

On 2014/08/07 07:58:24, Noel Gordon wrote:
> Check you have done all steps listed here ...
> 
>  
>
http://www.chromium.org/developers/how-tos/get-the-code#TOC-Working-with-repo...

Thanks!