Extend callingconv_case_by_case test to exercise the "pnaclcall"

Extend callingconv_case_by_case test to exercise the "pnaclcall" attribute (both via declarations, and via function pointer attribs).

Also turn on the nontrivial class test, now that we have done an upstream pulldown.

BUG= http://code.google.com/p/nativeclient/issues/detail?id=2274
BUG= http://code.google.com/p/nativeclient/issues/detail?id=2403

TEST= run_return_structs_fp_test && run_return_structs_decl_test &&
run_call_structs_fp_test && run_call_structs_decl_test

Committed: http://src.chromium.org/viewvc/native_client?view=rev&revision=7289

[jvoung - send to chromium..., 2011-11-16 19:09:06]

PTAL when you have spare cycles -- this version of the test should pass now

[robertm, 2011-11-16 20:38:28]

Some minor readability nits for the python code

http://codereview.chromium.org/8502006/diff/3001/tests/callingconv_case_by_ca...
File tests/callingconv_case_by_case/nacl.scons (right):

http://codereview.chromium.org/8502006/diff/3001/tests/callingconv_case_by_ca...
tests/callingconv_case_by_case/nacl.scons:24: def AddCommonFlags(envlist):
I think it would be better to sort of inline this, maybe -Wno-long-long could be
added to some base environment

http://codereview.chromium.org/8502006/diff/3001/tests/callingconv_case_by_ca...
tests/callingconv_case_by_case/nacl.scons:29: def MakeSelfTestEnv(base_env,
extra_flags):
more explanation for how this is used would not hurt, e.g.
"all envs are essentially the same, used for self consistency"

http://codereview.chromium.org/8502006/diff/3001/tests/callingconv_case_by_ca...
tests/callingconv_case_by_case/nacl.scons:90: # Helper function to make test
nodes.
it might be simpler to not inline this one

[jvoung - send to chromium..., 2011-11-17 23:06:01]

http://codereview.chromium.org/8502006/diff/3001/tests/callingconv_case_by_ca...
File tests/callingconv_case_by_case/nacl.scons (right):

http://codereview.chromium.org/8502006/diff/3001/tests/callingconv_case_by_ca...
tests/callingconv_case_by_case/nacl.scons:24: def AddCommonFlags(envlist):
On 2011/11/16 20:38:28, robertm wrote:
> I think it would be better to sort of inline this, maybe -Wno-long-long could
be
> added to some base environment

Well there's the part where it needs to give each of the 4 modules a different
-DMODULE number, which I would rather not inline / have to repeat everywhere.

http://codereview.chromium.org/8502006/diff/3001/tests/callingconv_case_by_ca...
tests/callingconv_case_by_case/nacl.scons:29: def MakeSelfTestEnv(base_env,
extra_flags):
On 2011/11/16 20:38:28, robertm wrote:
> more explanation for how this is used would not hurt, e.g.
> "all envs are essentially the same, used for self consistency"

Done.

http://codereview.chromium.org/8502006/diff/3001/tests/callingconv_case_by_ca...
tests/callingconv_case_by_case/nacl.scons:90: # Helper function to make test
nodes.
On 2011/11/16 20:38:28, robertm wrote:
> it might be simpler to not inline this one

Done (you mean to inline this one).

[jvoung - send to chromium..., 2011-11-28 18:30:35]

ptal

[robertm, 2011-11-28 18:49:45]

One small nit and one bigger remark:

Since we are already using c++ can some of the macro
magic be replaces with templates?

http://codereview.chromium.org/8502006/diff/12002/tests/callingconv_case_by_c...
File tests/callingconv_case_by_case/nacl.scons (right):

http://codereview.chromium.org/8502006/diff/12002/tests/callingconv_case_by_c...
tests/callingconv_case_by_case/nacl.scons:34: envlist = []
a list comprension would be more concise

[jvoung - send to chromium..., 2011-11-28 19:23:15]

It may be possible to use templates for some parts, but I don't think it will
work with everything. The macros were used to

- include or exclude function attributes at function decl
- include or exclude function pointer casts
- split up the code into different compilation units from one file.
- initialize the types appropriately (variable #sub-element types)
- check that the type values were preserved appropriately (variable #sub-element
types)
- give the test functions unique names / the right type declaration.

http://codereview.chromium.org/8502006/diff/12002/tests/callingconv_case_by_c...
File tests/callingconv_case_by_case/nacl.scons (right):

http://codereview.chromium.org/8502006/diff/12002/tests/callingconv_case_by_c...
tests/callingconv_case_by_case/nacl.scons:34: envlist = []
On 2011/11/28 18:49:45, robertm wrote:
> a list comprension would be more concise

Done.

[robertm, 2011-11-28 19:24:08]

LGTM


On 2011/11/28 19:23:15, jvoung wrote:
> It may be possible to use templates for some parts, but I don't think it will
> work with everything. The macros were used to
> 
> - include or exclude function attributes at function decl
> - include or exclude function pointer casts
> - split up the code into different compilation units from one file.
> - initialize the types appropriately (variable #sub-element types)
> - check that the type values were preserved appropriately (variable
#sub-element
> types)
> - give the test functions unique names / the right type declaration.
> 
>
http://codereview.chromium.org/8502006/diff/12002/tests/callingconv_case_by_c...
> File tests/callingconv_case_by_case/nacl.scons (right):
> 
>
http://codereview.chromium.org/8502006/diff/12002/tests/callingconv_case_by_c...
> tests/callingconv_case_by_case/nacl.scons:34: envlist = []
> On 2011/11/28 18:49:45, robertm wrote:
> > a list comprension would be more concise
> 
> Done.