Allow building mojo_shell and non-graphical apps/services on a Mac

Allow building mojo_shell and non-graphical apps/services on a Mac

The goal of this patch is to allow hacking on some pieces of the system
from a Mac, such as the generated bindings code and simple examples. It
is not the goal of this patch to make everything (or even most things)
work, or to support this as a development platform in general, just to
get the easy things going so folks can do some useful work from their
laptops on buses and whatnot.

This fixes up several compilation and link errors that show up building
on a Mac host and makes things that don't compile conditional in the
build system. mojo_shell and many of our unit test binaries run with
this patch if the current working directory is the out dir, but they're
missing some piece of rpath magic in order to run from anywhere.
mojob.py test doesn't do anything on a mac yet (and it's not clear what
it should do).

None of the graphics system is compiled in with this patch primiarily
for two reasons:
*) Lots of link errors in ui/ for events related code
*) Some link errors for CGL symbols as we don't appear to have the
   right frameworks configured in the GPU targets.

Depends on https://codereview.chromium.org/1313973005 to fix a linker
error in dart about having zero source files.

R=kulakowski@chromium.org, viettrungluu@chromium.org

Committed: https://chromium.googlesource.com/external/mojo/+/04362acf2200aabb2280d5fe11b2c17665cd89dd

[jamesr, 2015-09-10 01:11:24]

jamesr@chromium.org changed reviewers:
+ cstout@chromium.org, tonyg@chromium.org, viettrungluu@gmail.com

[jamesr, 2015-09-10 01:13:28]

jamesr@chromium.org changed reviewers:
+ cutchccutchan@google.com

[jamesr, 2015-09-10 01:13:29]

https://codereview.chromium.org/1327033004/diff/1/BUILD.gn
File BUILD.gn (right):

https://codereview.chromium.org/1327033004/diff/1/BUILD.gn#newcode30
BUILD.gn:30: if (is_linux || is_android) {
another way to do this would be to leave the normal list along and -= things in
a "if (is_mac)" clause. since is_mac is the "weird" case that makes a bit more
sense logically, but it also makes the file longer and easier to break by
removing or renaming something in the general list and forgetting to update the
is_mac section.  i'd be OK with either technique

https://codereview.chromium.org/1327033004/diff/1/mojo/dart/embedder/io/inter...
File mojo/dart/embedder/io/internet_address_posix.cc (right):

https://codereview.chromium.org/1327033004/diff/1/mojo/dart/embedder/io/inter...
mojo/dart/embedder/io/internet_address_posix.cc:17: intptr_t addr_length) {
according to 'diff' internet_address_linux.cc and internet_address_android.cc
were completely identical and both use only POSIX apis so i merged them.  this
diff is purely the result of clang-format

[kulakowski, 2015-09-10 15:58:39]

kulakowski@chromium.org changed reviewers:
+ kulakowski@chromium.org

[kulakowski, 2015-09-10 15:58:40]

https://codereview.chromium.org/1327033004/diff/1/examples/BUILD.gn
File examples/BUILD.gn (right):

https://codereview.chromium.org/1327033004/diff/1/examples/BUILD.gn#newcode51
examples/BUILD.gn:51: "//examples/window_manager",
nit: alphabetical order

[jamesr, 2015-09-10 17:18:27]

D'oh, I always forget that 'gn format' doesn't do that.

PTAL - the dart change has landed
(https://github.com/dart-lang/sdk/commit/38ab14f73428cb8252e26183046693299f2ff41b)
but not rolled.

[kulakowski, 2015-09-10 18:08:49]

lgtm

https://codereview.chromium.org/1327033004/diff/1/BUILD.gn
File BUILD.gn (right):

https://codereview.chromium.org/1327033004/diff/1/BUILD.gn#newcode30
BUILD.gn:30: if (is_linux || is_android) {
On 2015/09/10 01:13:29, jamesr wrote:
> another way to do this would be to leave the normal list along and -= things
in
> a "if (is_mac)" clause. since is_mac is the "weird" case that makes a bit more
> sense logically, but it also makes the file longer and easier to break by
> removing or renaming something in the general list and forgetting to update
the
> is_mac section.  i'd be OK with either technique

I'd find additions easier to reason about / track down than subtractive stuff.

[viettrungluu, 2015-09-10 18:11:08]

viettrungluu@chromium.org changed reviewers:
+ viettrungluu@chromium.org

[viettrungluu, 2015-09-10 18:11:09]

Probably you should also make pingable_app_dummy_empty.cc non-empty with a
comment explaining the reason for its existence.

https://codereview.chromium.org/1327033004/diff/20001/services/view_manager/d...
File services/view_manager/display_manager.cc (left):

https://codereview.chromium.org/1327033004/diff/20001/services/view_manager/d...
services/view_manager/display_manager.cc:20: using mojo::Rect;
Why?

https://codereview.chromium.org/1327033004/diff/20001/services/view_manager/v...
File services/view_manager/view_manager_service_impl.cc (left):

https://codereview.chromium.org/1327033004/diff/20001/services/view_manager/v...
services/view_manager/view_manager_service_impl.cc:24: using mojo::Rect;
"

https://codereview.chromium.org/1327033004/diff/20001/shell/application_manag...
File shell/application_manager/local_fetcher.cc (right):

https://codereview.chromium.org/1327033004/diff/20001/shell/application_manag...
shell/application_manager/local_fetcher.cc:62: #if defined(OS_MACOSX)
You should include build/build_config.h.

https://codereview.chromium.org/1327033004/diff/20001/shell/application_manag...
shell/application_manager/local_fetcher.cc:66: base::stat_wrapper_t stat_result;
Probably we should just use "struct stat64" here, or use base::stat_wrapper_t
above also.

https://codereview.chromium.org/1327033004/diff/20001/shell/application_manag...
File shell/application_manager/network_fetcher.cc (right):

https://codereview.chromium.org/1327033004/diff/20001/shell/application_manag...
shell/application_manager/network_fetcher.cc:28: #if defined(OS_LINUX)
Oops, we should have included build/build_config.h.

https://codereview.chromium.org/1327033004/diff/20001/shell/context.cc
File shell/context.cc (right):

https://codereview.chromium.org/1327033004/diff/20001/shell/context.cc#newcode43
shell/context.cc:43: #if !defined(OS_MACOSX)
You should include build/build_config.h.

[jamesr, 2015-09-10 19:59:56]

On 2015/09/10 at 18:11:09, viettrungluu wrote:
>
https://codereview.chromium.org/1327033004/diff/20001/services/view_manager/d...
> services/view_manager/display_manager.cc:20: using mojo::Rect;
> Why?

Some very common Mac headers define things called 'Rect' and 'Size' in the
global namespace.  This is dumb, but they do.  What this file tries to do is put
'mojo::Rect' in the global namespace, but this collides.

IMO putting 'mojo::Rect' in the global namespace isn't really a good idea in
this file anyway since this code deals with gfx::Rect a good bit as well so raw
'Rect' is ambiguous.

[viettrungluu, 2015-09-14 16:35:33]

lgtm w/nit

https://codereview.chromium.org/1327033004/diff/20001/shell/context.cc
File shell/context.cc (right):

https://codereview.chromium.org/1327033004/diff/20001/shell/context.cc#newcode43
shell/context.cc:43: #if !defined(OS_MACOSX)
On 2015/09/10 18:11:09, viettrungluu wrote:
> You should include build/build_config.h.

You forgot this file.

[jamesr, 2015-09-19 01:23:59]

Committed patchset #3 (id:40001) manually as
04362acf2200aabb2280d5fe11b2c17665cd89dd (presubmit successful).