iOS: Remove MB and Goma for GYP builders.

iOS: Remove MB and Goma for GYP builders.

This reverts "iOS: Add mb_type config for GYP builders."
and also removes Goma from these builders, as it will not work for WebRTC
since MB's GYP mode is hardcoded to use gyp_chromium.
MB should be working for GN since it's a stand-alone tool.

Goma doesn't work since $(goma_dir) in the JSON throws an error during
compile. So let's turn off Goma for these bots for now.

BUG=chromium:498746
NOTRY=True
TBR=smut@google.com

Committed: https://crrev.com/d3d8b6720a8d02282f3023c60f5e8bcfdad12555
Cr-Commit-Position: refs/heads/master@{#11578}

[kjellander_webrtc, 2016-02-11 14:39:40]

The CQ bit was checked by kjellander@webrtc.org

[commit-bot: I haz the power, 2016-02-11 14:39:50]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1685683007/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1685683007/1

[commit-bot: I haz the power, 2016-02-11 14:44:48]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2016-02-11 14:48:46]

Description was changed from

==========
iOS: Remove MB and Goma for GYP builders.

This reverts "iOS: Add mb_type config for GYP builders."
and also removes Goma from these builders, as it will not work for WebRTC
since MB's GYP mode is hardcoded to use gyp_chromium.
MB should be working for GN since it's a stand-alone tool.

Goma doesn't work since $(goma_dir) in the JSON throws an error during
compile. So let's turn off Goma for these bots for now.

BUG=chromium:498746
NOTRY=True
TBR=smut@google.com
==========

to

==========
iOS: Remove MB and Goma for GYP builders.

This reverts "iOS: Add mb_type config for GYP builders."
and also removes Goma from these builders, as it will not work for WebRTC
since MB's GYP mode is hardcoded to use gyp_chromium.
MB should be working for GN since it's a stand-alone tool.

Goma doesn't work since $(goma_dir) in the JSON throws an error during
compile. So let's turn off Goma for these bots for now.

BUG=chromium:498746
NOTRY=True
TBR=smut@google.com

Committed: https://crrev.com/d3d8b6720a8d02282f3023c60f5e8bcfdad12555
Cr-Commit-Position: refs/heads/master@{#11578}
==========

[commit-bot: I haz the power, 2016-02-11 14:48:47]

Patchset 1 (id:??) landed as
https://crrev.com/d3d8b6720a8d02282f3023c60f5e8bcfdad12555
Cr-Commit-Position: refs/heads/master@{#11578}

[kjellander_webrtc, 2016-02-23 20:04:53]

Can you explain why Goma doesn't work for us? We'd like to have it to speed
things up.

[smut, 2016-02-23 20:12:43]

I think we are not doing the Goma-related setup anywhere. +cc dpranke

[kjellander_webrtc, 2016-02-23 20:27:55]

On 2016/02/23 20:12:43, smut wrote:
> I think we are not doing the Goma-related setup anywhere. +cc dpranke

Do you mean the code at
https://code.google.com/p/chromium/codesearch#chromium/build/scripts/slave/re...
?

I digged up the old failure (the bots are moved now so the old links didn't
work).
The bot had environment like this during runhooks:
 GYP_DEFINES: chromium_ios_signing=0 target_arch=x64 use_goma=1
gomadir=$(goma_dir) component=static_library OS=ios

and failed on the compile step like this:

ninja -C /b/build/slave/mac64/build/src/out/Debug-iphonesimulator All -j80
ninja: Entering directory
`/b/build/slave/mac64/build/src/out/Debug-iphonesimulator'
ninja: error: build.ninja:1: bad $-escape (literal $ must be written as $$)
cc = ../../$(goma_dir)/gomacc ../../third_party/llvm-build/Release+Asser...
           ^ near here

Is the variable supposed to be expanded somewhere, or is it only working if MB
is used?

[Dirk Pranke, 2016-02-23 20:48:52]

On 2016/02/23 20:27:55, kjellander (webrtc) wrote:
> On 2016/02/23 20:12:43, smut wrote:
> > I think we are not doing the Goma-related setup anywhere. +cc dpranke
> 
> Do you mean the code at
>
https://code.google.com/p/chromium/codesearch#chromium/build/scripts/slave/re...
> ?
> 
> I digged up the old failure (the bots are moved now so the old links didn't
> work).
> The bot had environment like this during runhooks:
>  GYP_DEFINES: chromium_ios_signing=0 target_arch=x64 use_goma=1
> gomadir=$(goma_dir) component=static_library OS=ios
> 
> and failed on the compile step like this:
> 
> ninja -C /b/build/slave/mac64/build/src/out/Debug-iphonesimulator All -j80
> ninja: Entering directory
> `/b/build/slave/mac64/build/src/out/Debug-iphonesimulator'
> ninja: error: build.ninja:1: bad $-escape (literal $ must be written as $$)
> cc = ../../$(goma_dir)/gomacc ../../third_party/llvm-build/Release+Asser...
>            ^ near here
> 
> Is the variable supposed to be expanded somewhere, or is it only working if MB
> is used?

The ninja ios bots in the chromium waterfalls do use goma. As you see, they call
into the chromium module's compile method, which calls into compile.py to do 
the set up.

And, yes, $(goma_dir) is expanded via MB, and you probably need to use MB
for this to work. There's nothing that intrinsically ties MB to chromium (I
don't think),
so we can make it work in your checkout if need be. If you have a chromium
checkout on this bot, it should be easy; if not, we can copy MB over somewhere
and make it work.

Or we can change the ios recipe to have some other way of configuring goma
directly
as well.

[kjellander_webrtc, 2016-02-23 21:55:16]

On 2016/02/23 20:48:52, Dirk Pranke wrote:
> On 2016/02/23 20:27:55, kjellander (webrtc) wrote:
> > On 2016/02/23 20:12:43, smut wrote:
> > > I think we are not doing the Goma-related setup anywhere. +cc dpranke
> > 
> > Do you mean the code at
> >
>
https://code.google.com/p/chromium/codesearch#chromium/build/scripts/slave/re...
> > ?
> > 
> > I digged up the old failure (the bots are moved now so the old links didn't
> > work).
> > The bot had environment like this during runhooks:
> >  GYP_DEFINES: chromium_ios_signing=0 target_arch=x64 use_goma=1
> > gomadir=$(goma_dir) component=static_library OS=ios
> > 
> > and failed on the compile step like this:
> > 
> > ninja -C /b/build/slave/mac64/build/src/out/Debug-iphonesimulator All -j80
> > ninja: Entering directory
> > `/b/build/slave/mac64/build/src/out/Debug-iphonesimulator'
> > ninja: error: build.ninja:1: bad $-escape (literal $ must be written as $$)
> > cc = ../../$(goma_dir)/gomacc ../../third_party/llvm-build/Release+Asser...
> >            ^ near here
> > 
> > Is the variable supposed to be expanded somewhere, or is it only working if
MB
> > is used?
> 
> The ninja ios bots in the chromium waterfalls do use goma. As you see, they
call
> into the chromium module's compile method, which calls into compile.py to do 
> the set up.
> 
> And, yes, $(goma_dir) is expanded via MB, and you probably need to use MB
> for this to work. There's nothing that intrinsically ties MB to chromium (I
> don't think),
> so we can make it work in your checkout if need be. If you have a chromium
> checkout on this bot, it should be easy; if not, we can copy MB over somewhere
> and make it work.
> 
> Or we can change the ios recipe to have some other way of configuring goma
> directly
> as well.

Right. I recall now that I figured out MB was the problem so I landed without
Goma because I wanted to get these bots up using the ios recipe module.
Since we have a Chromium copy in WebRTC we can easily add tools/mb to the build.
It would be nice though if we could keep our own JSON files for MB in the WebRTC
repo (i.e. not in tools/mb). 
Otherwise we'd have to do a roll every time we do a change in them. We would
also not be able to run trybots with changes in the JSON files, which is one of
the main benefits with MB once the GN switch is moving closer. How would we do
that? Should I file a feature request against MB for that case?

[Dirk Pranke, 2016-02-23 22:14:02]

On 2016/02/23 21:55:16, kjellander (webrtc) wrote:
> It would be nice though if we could keep our own JSON files for MB in the
WebRTC
> repo (i.e. not in tools/mb). 

MB supports this already; pass the path you want to use in the mb_config_path
parameter to 
m.chromium.run_mb().

Note that the config files need to be self-contained (you won't be able to
reference the existing
definitions in the chromium file), so copy over everything you need, but that
should be straightforward.

(And the files are .pyl , not .json ;).