Rename is_chromeos.

Rename is_chromeos.

is_chromeos does a poor job of distinguishing developer builds of
chrome for chromeos for your desktop, and chrome for chromeos on
a device (where the runtimes are non local, for example)

Split this into is_chromeos_ui and is_chromeos_os.

BUG=519943
TEST=ChromiumOS compiles equivalently.

Defered until face-to-face time allows.
Follow on changes may be unfrozen enar the same time :
https://codereview.chromium.org/1291703008
https://codereview.chromium.org/1290663005/
https://codereview.chromium.org/1286273004/
https://codereview.chromium.org/1290033003/
https://codereview.chromium.org/1291703007/
https://codereview.chromium.org/1287853007/
and
https://codereview.chromium.org/1285243004/

[Peter Mayo, 2015-08-13 01:27:23]

Split this into the mechanical an business part CLs since we have to have
backwards compatability for separate repos.

[Peter Mayo, 2015-08-13 01:46:46]

https://codereview.chromium.org/1283313004/diff/60001/build/config/BUILDCONFI...
File build/config/BUILDCONFIG.gn (right):

https://codereview.chromium.org/1283313004/diff/60001/build/config/BUILDCONFI...
build/config/BUILDCONFIG.gn:154: cros_use_custom_toolchain = false
This change can be separated.

https://codereview.chromium.org/1283313004/diff/60001/build/config/BUILDCONFI...
build/config/BUILDCONFIG.gn:207: assert(is_chromeos_ui)
Provide message with assert.

https://codereview.chromium.org/1283313004/diff/60001/build/config/BUILDCONFI...
build/config/BUILDCONFIG.gn:596: if (cros_use_custom_toolchain) {
Part of second change.

[Peter Mayo, 2015-08-13 01:52:02]

Further trim.

[Peter Mayo, 2015-08-13 02:03:25]

On 2015/08/13 01:27:23, Peter Mayo wrote:
> Split this into the mechanical an business part CLs since we have to have
> backwards compatability for separate repos.

Done: https://codereview.chromium.org/1291703008/

[Peter Mayo, 2015-08-13 02:07:38]

https://codereview.chromium.org/1283313004/diff/60001/build/config/BUILDCONFI...
File build/config/BUILDCONFIG.gn (right):

https://codereview.chromium.org/1283313004/diff/60001/build/config/BUILDCONFI...
build/config/BUILDCONFIG.gn:154: cros_use_custom_toolchain = false
On 2015/08/13 01:46:46, Peter Mayo wrote:
> This change can be separated.

Done.  https://codereview.chromium.org/1287153002/

https://codereview.chromium.org/1283313004/diff/60001/build/config/BUILDCONFI...
build/config/BUILDCONFIG.gn:207: assert(is_chromeos_ui)
On 2015/08/13 01:46:46, Peter Mayo wrote:
> Provide message with assert.

Done.

[Peter Mayo, 2015-08-13 02:10:24]

Toggle desktop off by default on ChromeOS

[Peter Mayo, 2015-08-13 02:12:12]

Add Todo

[Peter Mayo, 2015-08-13 02:13:55]

Add Todo

[Peter Mayo, 2015-08-13 02:14:03]

Patchset #8 (id:140001) has been deleted

[Peter Mayo, 2015-08-13 02:17:23]

petermayo@chromium.org changed reviewers:
+ brettw@chromium.org, dpranke@chromium.org

[Peter Mayo, 2015-08-13 02:17:23]

PTAL.

[brettw, 2015-08-13 19:55:49]

https://codereview.chromium.org/1283313004/diff/160001/build/config/BUILDCONF...
File build/config/BUILDCONFIG.gn (right):

https://codereview.chromium.org/1283313004/diff/160001/build/config/BUILDCONF...
build/config/BUILDCONFIG.gn:129: is_chromeos_ui = current_os == "chromeos"
Given the below definition for is_chromeos, having this here is unavoidable for
now. However, I don't think this should be in the mast build config file.

Most stuff like this is in build/config/ui.gni. However, these are things that
are global for legacy reasons. Since this separation is new, we should do it
correctly and put it in //ui/config.gni so it can be used from UI and below, but
not e.g. net and base. Things like net and base would use only is_chromeos_os.

We could land it like this and then fix the callers and then delete the
is_chromeos_ui thing.

I think a way to land this I might prefer would be to add an is_chromeos_ui flag
to ui/config.gni that's defined to be is_chromeos for now. Update callers to use
this flag and import file, then separate the meaning of the two flags. This
approach avoids the intermediate mess in BUILDCONFIG while I think being
equivalent (I'm trying REALLY hard to keep this file clean to avoid the
common.gypi problem).

https://codereview.chromium.org/1283313004/diff/160001/build/config/BUILDCONF...
build/config/BUILDCONFIG.gn:168: # - is_android, is_chromeos_os, is_ios, and
is_win should be obvious.
In this case, is_chromeos_os isn't so obvious, I think this should explicitly
mention the _ui and _os flags and how they differ.

[Peter Mayo, 2015-08-14 18:08:47]

On 2015/08/13 19:55:49, brettw wrote:
>
https://codereview.chromium.org/1283313004/diff/160001/build/config/BUILDCONF...
> File build/config/BUILDCONFIG.gn (right):
> 
>
https://codereview.chromium.org/1283313004/diff/160001/build/config/BUILDCONF...
> build/config/BUILDCONFIG.gn:129: is_chromeos_ui = current_os == "chromeos"
> Given the below definition for is_chromeos, having this here is unavoidable
for
> now. However, I don't think this should be in the mast build config file.
> 
> Most stuff like this is in build/config/ui.gni. However, these are things that
> are global for legacy reasons.

The tendrils of confusion spread wide for what the "ui" in this means.  I am
currently thinking of this as meaning "the UI for the OS ChromeOS", rather than
"the UI for the browser on ChromeOS".  There are a lot of usefully stubbed out
pieces to emulate a ChromeOS environment, (e.g in base we remove a power_manager
piece that goes everywhere else posix).  I considered just calling this building
the interface, but is_chromeos_if is a bad name, and chetting it too long is
also bad given how frequently it is (currently) referenced.

This flag also controls things like the unittests, which run on linux under
emulation because the services that the frameworks rely on don't necessarily
make sense in an embedded world. (Let's not diverge too far into unittesting
theory)

> Since this separation is new, we should do it
> correctly and put it in //ui/config.gni so it can be used from UI and below,
but
> not e.g. net and base. Things like net and base would use only is_chromeos_os.

As I pointed out above, I'm not sure net and base can usefully use only
is_chromeos_os

Looking at https://codereview.chromium.org/1291703008/diff/1/net/BUILD.gn I see
that the _ui flag is more useful to turn off proxy_configs, cert_unittests, and
gnome IO.  Testing that the version of linux code without these still work is
useful.


> 
> We could land it like this and then fix the callers and then delete the
> is_chromeos_ui thing.

> 
> I think a way to land this I might prefer would be to add an is_chromeos_ui
flag
> to ui/config.gni that's defined to be is_chromeos for now. Update callers to
use
> this flag and import file, then separate the meaning of the two flags. This
> approach avoids the intermediate mess in BUILDCONFIG while I think being
> equivalent (I'm trying REALLY hard to keep this file clean to avoid the
> common.gypi problem).

Another way I can think of to keep this cleaner is to put the
multiplicity/subtlety of chromeos in a chromeos specific file that is only
loaded when a single global is set (either through args, or OS)  The the things
that are not clear from the top level (i.e. need to specify a proper subset of
the chromeos-sy builds) can depend on the clarification file, guarded by a
single global.

> 
>
https://codereview.chromium.org/1283313004/diff/160001/build/config/BUILDCONF...
> build/config/BUILDCONFIG.gn:168: # - is_android, is_chromeos_os, is_ios, and
> is_win should be obvious.
> In this case, is_chromeos_os isn't so obvious, I think this should explicitly
> mention the _ui and _os flags and how they differ.

Ack - let's deal with the other issue first though.

[Peter Mayo, 2015-08-14 19:10:13]

On 2015/08/14 18:08:47, Peter Mayo wrote:
> .. and chetting it too long is
> also bad ..
... and letting it get too long is bad ...

<shakes head while scratching it>

[ddorwin, 2015-08-18 19:02:12]

(Drive-by. I arrived here from https://codereview.chromium.org/1290663005/.)

I'm not sure "ui" is an accurate label for such all code controlled by that
flag. There are certain non-UI-related features/capabilities/behaviors that
Chrome OS has that we would want to trigger on your desktop. "behavior" seems
more accurate.

Also, would "_device" be more accurate than "_os"?

[Peter Mayo, 2015-08-18 19:54:14]

https://codereview.chromium.org/1283313004/diff/160001/build/config/BUILDCONF...
File build/config/BUILDCONFIG.gn (right):

https://codereview.chromium.org/1283313004/diff/160001/build/config/BUILDCONF...
build/config/BUILDCONFIG.gn:340: if (!is_chromeos_ui) {
Is it useful to make this filter only available from UI and below?

[Peter Mayo, 2015-08-18 19:59:17]

On 2015/08/18 19:02:12, ddorwin wrote:
> (Drive-by. I arrived here from https://codereview.chromium.org/1290663005/.)
> 
> I'm not sure "ui" is an accurate label for such all code controlled by that
> flag. There are certain non-UI-related features/capabilities/behaviors that
> Chrome OS has that we would want to trigger on your desktop. "behavior" seems
> more accurate.

I'm pretty sure UI is not a good label, but I don't think behavior is net
better.
It doesn't convey the compiling and linking of new pieces to me, and is quite
long
to read.

Any other creatives?

> 
> Also, would "_device" be more accurate than "_os"?

Not in my opinion.
I think device is a good mnemonic for what we are thinking about mostly, but
looking
at two hinged all-in-one devices with solid state storage and calling one a
device
and the other not a device seems peculiar.  The critical difference is how/what
it boots.

[Peter Mayo, 2015-08-18 22:12:55]

On 2015/08/18 19:59:17, Peter Mayo wrote:
> On 2015/08/18 19:02:12, ddorwin wrote:
> > (Drive-by. I arrived here from https://codereview.chromium.org/1290663005/.)
> > 
> > I'm not sure "ui" is an accurate label for such all code controlled by that
> > flag. There are certain non-UI-related features/capabilities/behaviors that
> > Chrome OS has that we would want to trigger on your desktop. "behavior"
seems
> > more accurate.
> 
> I'm pretty sure UI is not a good label, but I don't think behavior is net
> better.
> It doesn't convey the compiling and linking of new pieces to me, and is quite
> long
> to read.
> 
> Any other creatives?
> 
> > 
> > Also, would "_device" be more accurate than "_os"?
> 
> Not in my opinion.
> I think device is a good mnemonic for what we are thinking about mostly, but
> looking
> at two hinged all-in-one devices with solid state storage and calling one a
> device
> and the other not a device seems peculiar.  The critical difference is
how/what
> it boots.

What do people think about using frontend and backend for the difference here?

fe/be are equally short.  They are grouped but keep the notion of it has all the
pieces to be X but sits on Y to get it done.

[Peter Mayo, 2015-08-18 22:16:29]

move more to ui.gni

[Peter Mayo, 2015-08-19 21:41:49]

Tweak use of flag in ui.gni

[Peter Mayo, 2015-08-19 21:42:05]

PTAL

https://codereview.chromium.org/1283313004/diff/180001/build/config/BUILDCONF...
File build/config/BUILDCONFIG.gn (right):

https://codereview.chromium.org/1283313004/diff/180001/build/config/BUILDCONF...
build/config/BUILDCONFIG.gn:343: #TODO(petermayo): Move this to ui.gni with the
removal of is_chromeos
This affects a lot of places.  North of 400 directories have some of this stuff.

https://codereview.chromium.org/1283313004/diff/200001/build/config/BUILDCONF...
File build/config/BUILDCONFIG.gn (right):

https://codereview.chromium.org/1283313004/diff/200001/build/config/BUILDCONF...
build/config/BUILDCONFIG.gn:160: cros_use_custom_toolchain = false
prefer https://codereview.chromium.org/1287153002/

https://codereview.chromium.org/1283313004/diff/200001/build/config/ui.gni
File build/config/ui.gni (right):

https://codereview.chromium.org/1283313004/diff/200001/build/config/ui.gni#ne...
build/config/ui.gni:36: toolkit_views = is_mac || is_win || is_chromeos ||
use_aura
oops, the import is here so we can rename this.

https://codereview.chromium.org/1283313004/diff/200001/build/config/ui.gni#ne...
build/config/ui.gni:75: enable_topchrome_md = is_chromeos || is_win || is_linux
oops, the import is here so we can rename this.

[Dirk Pranke, 2015-08-19 22:05:50]

I defer this to Brett. I might prefer 'use_cros_frontend' to 'use_cros_fe'.

[Peter Mayo, 2015-08-19 22:10:57]

petermayo@chromium.org changed reviewers:
- dpranke@chromium.org

[Peter Mayo, 2015-08-20 14:35:01]

Patchset #1 (id:1) has been deleted