Re: getting EBADF on MSVC

[Bruno Haible]

Bastien ROUCARIES wrote:
> why not
>  #else
>  # define TRY_MSVC_INVAL if (1)
>  # define CATCH_MSVC_INVAL else do
>  # define DONE_MSVC_INVAL while(0)
>  #endif

This would be needed if we wanted the user to write

   DONE_MSVC_INVAL;

instead of

   DONE_MSVC_INVAL

But what's the point of that semicolon? The block does not look like an
expression.

> why not something like:
> #if HAVE_DUP2
> # undef dup2
> # if MSVC_HACK_NEEDED              /* NEW */
> #  include "msvc-hack.h"           /* NEW */
> static int dup2_msvcsafe(int fd, int desired_fd)
> {
> 
> }
> # else
> static inline dup2_msvcsafe(int fd, int desired_fd)
> {
>  return dup2();
> }
> # endif                            /* NEW */
> 
> int rpl_dup2 (int fd, int desired_fd)
>    {
>        ...
>        result = dup2 (fd, desired_fd);
>        ...
>     }

Is this code (with a #if and two extra functions) really easier to understand
than what we have in lib/dup2.c now?

> Moreover _get_osfhandle (fd) should be render safe globally.

Yes, that might well be necessary.

> May be we should create a <io.h> module ?

The users of gnulib are not meant to include a file <io.h>, since that's not
in POSIX. So, a gnulib-private header file and a gnulib-private function will
do. Maybe I will call them "io-nothrow.h" and _get_osfhandle_nothrow().

Bruno
-- 
In memoriam James A. Garfield <http://en.wikipedia.org/wiki/James_A._Garfield>