[NaCl SDK] Make the SDK examples buildable as a packaged app.

[NaCl SDK] Make the SDK examples buildable as a packaged app.

BUG=169313
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=193157

[noelallen1, 2013-04-08 17:42:56]

A few questions, and hello_world_instance3d comment seems wrong, should be more
like hello_world_stdio I believe.

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
File native_client_sdk/src/build_tools/generate_make.py (right):

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
native_client_sdk/src/build_tools/generate_make.py:109: 'TYPE': (str, ['main',
'lib', 'so', 'so-standalone'], True),
We should document somewhere what these different types mean.

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
native_client_sdk/src/build_tools/generate_make.py:266: configs = [c.lower() for
c in configs]
This change assumes it's always correct for windows.  Do you know that's true?

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
native_client_sdk/src/build_tools/generate_make.py:372: if first_toolchain:
Is this actually what we want?  Do we really want 'First', or do we want 'Only'
such that examples that don't support that toolchain get skipped?

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
native_client_sdk/src/build_tools/generate_make.py:417:
parser.add_option('--first-valid-toolchain',
Same comment as above.  Is this a testing optimization for development?

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/exa...
File native_client_sdk/src/examples/hello_world_instance3d/index.html (right):

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/exa...
native_client_sdk/src/examples/hello_world_instance3d/index.html:17: <p>The
Hello World GLES 2.0 example demonstrates how to create a 3D cube that
This is actually using C++, PPAPIinstance, and nacl_io right?

[binji, 2013-04-08 19:33:52]

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
File native_client_sdk/src/build_tools/generate_make.py (right):

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
native_client_sdk/src/build_tools/generate_make.py:109: 'TYPE': (str, ['main',
'lib', 'so', 'so-standalone'], True),
On 2013/04/08 17:42:56, noelallen1 wrote:
> We should document somewhere what these different types mean.

Done.

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
native_client_sdk/src/build_tools/generate_make.py:266: configs = [c.lower() for
c in configs]
On 2013/04/08 17:42:56, noelallen1 wrote:
> This change assumes it's always correct for windows.  Do you know that's
that's true?

Below we validate that config is in ['Debug', 'Release']. This code should be
logically the same as before -- the difference is that we can pass in just
['Release'] to config, to only build the release builds.

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
native_client_sdk/src/build_tools/generate_make.py:372: if first_toolchain:
On 2013/04/08 17:42:56, noelallen1 wrote:
> Is this actually what we want?  Do we really want 'First', or do we want
'Only'
> such that examples that don't support that toolchain get skipped?

This change is so that we don't include extra toolchain builds for the packaged
app (reduces the app size from 180M -> 10M zipped)

the trouble is, we can't build everything as newlib (dlopen wouldn't work) or
glibc (arm wouldn't work) -- so this says, build the first toolchain that is
valid for the example. Usually that is newlib, but for dlopen it is glibc.

Maybe there's a better way of doing this...?

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/exa...
File native_client_sdk/src/examples/hello_world_instance3d/index.html (right):

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/exa...
native_client_sdk/src/examples/hello_world_instance3d/index.html:17: <p>The
Hello World GLES 2.0 example demonstrates how to create a 3D cube that
On 2013/04/08 17:42:56, noelallen1 wrote:
> This is actually using C++, PPAPIinstance, and nacl_io right?

Good catch, I copied this from the .dsc file... :)

[noelallen1, 2013-04-08 19:48:16]

lgtm

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
File native_client_sdk/src/build_tools/generate_make.py (right):

https://codereview.chromium.org/13488007/diff/10001/native_client_sdk/src/bui...
native_client_sdk/src/build_tools/generate_make.py:372: if first_toolchain:
That's fine, I though you were trying to solve a different problem.

On 2013/04/08 19:33:52, binji wrote:
> On 2013/04/08 17:42:56, noelallen1 wrote:
> > Is this actually what we want?  Do we really want 'First', or do we want
> 'Only'
> > such that examples that don't support that toolchain get skipped?
> 
> This change is so that we don't include extra toolchain builds for the
packaged
> app (reduces the app size from 180M -> 10M zipped)
> 
> the trouble is, we can't build everything as newlib (dlopen wouldn't work) or
> glibc (arm wouldn't work) -- so this says, build the first toolchain that is
> valid for the example. Usually that is newlib, but for dlopen it is glibc.
> 
> Maybe there's a better way of doing this...?

[binji, 2013-04-09 18:24:37]

The linux_nacl_sdk trybot is currently broken (see crbug.com/229470), but I've
run locally and it builds fine. I'll watch the tree afterward.

[binji, 2013-04-09 18:36:47]

Committed patchset #8 manually as r193157 (presubmit successful).