Make a gyp file for the pnacl_irt_shim.

Make a gyp file for the pnacl_irt_shim (nacl-side)

- Add generated_sources variable to untrusted.gypi to handle
  sources that are generated and thus unscannable.

BUG= http://code.google.com/p/nativeclient/issues/detail?id=2465
TEST= manual

- gclient runhooks
- run gyp make
- check output directory (out/Release/obj/gen/tc_newlib/lib64/)
- touch generator python files
- run make again to check that it re-ran the generator

[noelallen_use_chromium, 2011-12-07 19:20:47]

You could add the step to generate 'C' style .h headers to SHARED_INTERMEDIATE. 
 Then depend on that step so that your shim step makes more sense.   (It only
depends on the header generator step and the *.py files)

The PPAPI headers should be getting moved to this new location so that would be
a step in the correct direction.

The issue is that generating the PPAPI headers (IMO) is a Chrome thing.  So that
should happen in the Chrome tree.   That means the SHIM is a Chrome thing and
should happen in the chrome tree.

Unfortunately, we do not have a good way of running those GYP pieces in NaCl
because of the tree hierarchy.   Which really means all this should run only in
Chrome.

Do we need to run this in NaCl?

[jvoung - send to chromium..., 2011-12-08 22:39:02]

On 2011/12/07 19:20:47, noelallen wrote:
> You could add the step to generate 'C' style .h headers to
SHARED_INTERMEDIATE. 
>  Then depend on that step so that your shim step makes more sense.   (It only
> depends on the header generator step and the *.py files)
> 

Started thinking about how to write the 'action' for generating the headers. The
'inputs' of the action will end up look similar (using gyp_glob.py), since the
headers depend the IDL files + python scripts. The 'outputs' may look a bit
ugly, since we will basically take the list of IDL files and map that list to a
list a headers.

Is there another way?


> The PPAPI headers should be getting moved to this new location so that would
be
> a step in the correct direction.
> 
> The issue is that generating the PPAPI headers (IMO) is a Chrome thing.  So
that
> should happen in the Chrome tree.   That means the SHIM is a Chrome thing and
> should happen in the chrome tree.
> 
> Unfortunately, we do not have a good way of running those GYP pieces in NaCl
> because of the tree hierarchy.   Which really means all this should run only
in
> Chrome.
> 
> Do we need to run this in NaCl?

As we discussed yesterday, there are some commandline toolchain tests that use
sel_universal to mock the file-handle-passing to the sandboxed compilers. These
toolchain tests will still need a shim file since it uses the linker same
"commandline" as it does in the browser, even though the shim file in NaCl may
be stale.

I ended up moving the pnacl_irt_shim.gyp file to the ppapi directory:

http://codereview.chromium.org/8873035/

To get rid of the "add_path_prefix.py" script/hack, since most of the files it
refers to are in the ppapi directory. It also seemed to help with building this
in Windows.

However, splitting up the CL means that I was only able to manually test this...