Hi
On 10 April 2016 at 22:30, Samuel Thibault <samuel.thiba...@gnu.org> wrote:
> Hello,
>
> Flavio Cruz, on Tue 05 Apr 2016 02:08:44 +0200, wrote:
> > --- a/include/mach/message.h
> > +++ b/include/mach/message.h
> > @@ -188,8 +188,8 @@ typedef struct {
> > * a mach_msg_type_long_t should be zero.
> > */
> >
> > -typedef unsigned int mach_msg_type_name_t;
> > -typedef unsigned int mach_msg_type_size_t;
> > +typedef unsigned char mach_msg_type_name_t;
> > +typedef unsigned short mach_msg_type_size_t;
>
> Looking at this again, I'm realizing that this would change the ABI
> of a big lot of functions, those which have a mach_msg_type_name_t *
> parameter. I don't think we can afford doing that :) Which warning do
> you get, exactly?
>
I fully understand your worries about changing the ABI :) However, it could
be nice to have a separate branch with ABI changes that could be
incorporated into a major release.
The warning I'm talking about is the following:
In file included from ipc/mach_port.server.c:10:0:
ipc/mach_port.server.c: In function '_Xmach_port_extract_right':
../gnumach/include/mach/message.h:268:35: warning: large integer implicitly
truncated to unsigned type [-Woverflow]
#define MACH_MSG_TYPE_POLYMORPHIC ((mach_msg_type_name_t) -1)
^
ipc/mach_port.server.c:1355:22: note: in expansion of macro
'MACH_MSG_TYPE_POLYMORPHIC'
/* msgt_name = */ MACH_MSG_TYPE_POLYMORPHIC,
^
This shows when compiling some stubs since now mig generates code with
MACH_MSG_TYPE_POLYMORPHIC instead of -1.
I think this can be trivially fixed by removing the cast in
MACH_MSG_TYPE_POLYMORPHIC.
>
> Samuel
>
--
Flávio Cruz / flavioc...@gmail.com
