This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: libstdc++ v3 wants more libio functions now


> From: "Geert Bosch" <bosch@gnat.com>
> To: "Michael Vance" <briareos@lokigames.com>, "Mike Stump" <mrs@windriver.com>
> Date: Wed, 26 Jul 2000 08:16:21 -0400
> Reply-To: "Geert Bosch" <bosch@gnat.com>

> On Tue, 25 Jul 2000 12:06:16 -0700, Michael Vance wrote:

>   Is wchar_t still defined as a four-byte quantity in GCC? If so, is
>   there a document or archived mlist discussion somewhere documenting
>   why this is? I know that the spec says it is
>   "implementation-dependent", but every implementation I'm aware of does
>   it as a 2-byte wide character. This was a nasty PITA when porting some
>   SDK code during Solder of Fortune... does GCC support some locale that
>   demands four-byte wide characters?

> I am just fixing the code in GNAT for interfacing with C wide
> characters.  GNAT assumed that C wchar_t was a 16-bit quantity, but
> we were notified by a user that this actually was defined as
> Integer'Size in GCC.

The user's comment is too simplistic for someone to code against.

The size can be found by compiling something like:

#include <stddef.h>
int main() { printf ("%d\n", sizeof(wchar_t)); }

in C for C++.  That will give either 2 or 4 on all systems I know
about (carefully excluding one that is totally broken).

Assuming you know what size it is, is wrong.

> What do the different Unices expect in their system calls?

Different answers.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]