Restructure the structs/unions involved in SRPC argument passing. This will...

Restructure the structs/unions involved in SRPC argument passing.  This will
make the serialization changes I am currently working on more robust.
BUG=none
TEST=all srpc tests


Committed: http://src.chromium.org/viewvc/native_client?view=rev&revision=3899

[sehr (please use chromium), 2010-12-05 16:58:34]

Cliff,

Here's what we discussed on Friday.  Please take a look.

David

[krasin, 2010-12-06 11:44:34]

http://codereview.chromium.org/5622003/diff/31003/src/shared/srpc/nacl_srpc.h
File src/shared/srpc/nacl_srpc.h (right):

http://codereview.chromium.org/5622003/diff/31003/src/shared/srpc/nacl_srpc.h...
src/shared/srpc/nacl_srpc.h:136: #define NACL_SRPC_ARG_SERIALIZED_FIELDS        
                               \
Why it's a macro? It's used once...

http://codereview.chromium.org/5622003/diff/31003/src/shared/srpc/nacl_srpc.h...
src/shared/srpc/nacl_srpc.h:222: NACL_SRPC_RPC_SERIALIZED_FIELDS;
Ditto.

[sehr (please use chromium), 2010-12-06 17:07:40]

Adding inline comments to my previous mail.

http://codereview.chromium.org/5622003/diff/31003/src/shared/srpc/nacl_srpc.h
File src/shared/srpc/nacl_srpc.h (right):

http://codereview.chromium.org/5622003/diff/31003/src/shared/srpc/nacl_srpc.h...
src/shared/srpc/nacl_srpc.h:136: #define NACL_SRPC_ARG_SERIALIZED_FIELDS        
                               \
On 2010/12/06 11:44:34, krasin wrote:
> Why it's a macro? It's used once...

I mailed the response, but this CL does not change the serialization.  Once
serialization is changed, another struct will contain a use of this macro.

http://codereview.chromium.org/5622003/diff/31003/src/shared/srpc/nacl_srpc.h...
src/shared/srpc/nacl_srpc.h:222: NACL_SRPC_RPC_SERIALIZED_FIELDS;
On 2010/12/06 11:44:34, krasin wrote:
> Ditto.

Ditto back :-)

[Cliff L. Biffle, 2010-12-08 21:51:51]

LGTM, not because I love giant defines, but because I think I see how this will
make your *other* CL cleaner.

http://codereview.chromium.org/5622003/diff/28002/src/shared/srpc/invoke.c
File src/shared/srpc/invoke.c (right):

http://codereview.chromium.org/5622003/diff/28002/src/shared/srpc/invoke.c#ne...
src/shared/srpc/invoke.c:153: (arg)->array_name = va_arg(va, impl_type)
If you wanted, you could have array_name be e.g. "lval" and use
(arg)->arrays.array_name.  But, not a big deal.

[sehr (please use chromium), 2010-12-08 23:25:12]

On 2010/12/08 21:51:51, Cliff L. Biffle wrote:
> LGTM, not because I love giant defines, but because I think I see how this
will
> make your *other* CL cleaner.
> 
> http://codereview.chromium.org/5622003/diff/28002/src/shared/srpc/invoke.c
> File src/shared/srpc/invoke.c (right):
> 
>
http://codereview.chromium.org/5622003/diff/28002/src/shared/srpc/invoke.c#ne...
> src/shared/srpc/invoke.c:153: (arg)->array_name = va_arg(va, impl_type)
> If you wanted, you could have array_name be e.g. "lval" and use
> (arg)->arrays.array_name.  But, not a big deal.

Thanks.  Committed as r3899.