Support pointers in the NaCl bindings generator.

Support pointers in the NaCl bindings generator.

Previously the generator assumed any Mojo interface with a pointer argument was
broken. Going forward, we'll be bringing up these interfaces one at a time, so
explicitly mark interfaces as broken rather than infering from the arguments.

BUG=401761
R=viettrungluu@chromium.org

Committed: https://chromium.googlesource.com/external/mojo/+/ed1b03fbbb457def1e39b2dea4e020299ad315b1

[Nick Bray (chromium), 2015-04-13 19:52:22]

ncbray@chromium.org changed reviewers:
+ vtl@google.com

[Nick Bray (chromium), 2015-04-13 19:52:22]

As warned, this CL is a little awkward, but I am unsure  how to cut the plumbing
up otherwise.

[viettrungluu, 2015-04-13 20:09:41]

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

[viettrungluu, 2015-04-13 20:09:42]

LGTM w/nits

https://codereview.chromium.org/1086743002/diff/1/nacl_bindings/mojo_syscall_...
File nacl_bindings/mojo_syscall_internal.h (right):

https://codereview.chromium.org/1086743002/diff/1/nacl_bindings/mojo_syscall_...
nacl_bindings/mojo_syscall_internal.h:54: bool ConvertPointerInput(struct
NaClApp* nap, uint32_t user_ptr, T** value) {
Note: You could achieve this via a specialization of ConvertScalarInput (modulo
the fact that the name would then be confusing).

But I agree that having separate functions is clearer, in which case having a
suitable static_assert in ConvertScalarInput() (namely, that T is not a pointer)
is probably a good idea.[*]

[*] Since NaCl support isn't building on Android yet[**], you can assume that
the C++11 std::is_pointer (from <type_traits>) is available.
[**] Hopefully, when we want to build on Android, we'll be using the C++11
standard library on Android too.

https://codereview.chromium.org/1086743002/diff/1/nacl_bindings/mojo_syscall_...
nacl_bindings/mojo_syscall_internal.h:70: bool ConvertPointerInOut(struct
NaClApp* nap,
"

https://codereview.chromium.org/1086743002/diff/1/nacl_bindings_generator/int...
File nacl_bindings_generator/interface.py (right):

https://codereview.chromium.org/1086743002/diff/1/nacl_bindings_generator/int...
nacl_bindings_generator/interface.py:30: f.IsBrokenInNaCl()
Maybe add suitable TODOs for these?

[Nick Bray (chromium), 2015-04-13 21:13:52]

https://codereview.chromium.org/1086743002/diff/1/nacl_bindings/mojo_syscall_...
File nacl_bindings/mojo_syscall_internal.h (right):

https://codereview.chromium.org/1086743002/diff/1/nacl_bindings/mojo_syscall_...
nacl_bindings/mojo_syscall_internal.h:54: bool ConvertPointerInput(struct
NaClApp* nap, uint32_t user_ptr, T** value) {
On 2015/04/13 20:09:42, viettrungluu wrote:
> Note: You could achieve this via a specialization of ConvertScalarInput
(modulo
> the fact that the name would then be confusing).
> 
> But I agree that having separate functions is clearer, in which case having a
> suitable static_assert in ConvertScalarInput() (namely, that T is not a
pointer)
> is probably a good idea.[*]
> 
> [*] Since NaCl support isn't building on Android yet[**], you can assume that
> the C++11 std::is_pointer (from <type_traits>) is available.
> [**] Hopefully, when we want to build on Android, we'll be using the C++11
> standard library on Android too.

Done.

https://codereview.chromium.org/1086743002/diff/1/nacl_bindings_generator/int...
File nacl_bindings_generator/interface.py (right):

https://codereview.chromium.org/1086743002/diff/1/nacl_bindings_generator/int...
nacl_bindings_generator/interface.py:30: f.IsBrokenInNaCl()
On 2015/04/13 20:09:42, viettrungluu wrote:
> Maybe add suitable TODOs for these?

Done.

[Nick Bray (chromium), 2015-04-13 21:23:02]

Committed patchset #2 (id:20001) manually as
ed1b03fbbb457def1e39b2dea4e020299ad315b1 (presubmit successful).