On Wed, Jun 30, 2004 at 09:12:12AM +0200, Ulf Lamping wrote:
> ethereal-dev@xxxxxxxxxxxx schrieb am 30.06.04 08:59:28:
> >
> > guy 2004/06/30 01:58:59 CDT
> >
> > Modified files:
> > gtk proto_draw.c rtp_analysis.c
> > Log:
> > On at least some platforms, a #define of O_BINARY is needed even if
> > <fcntl.h> is included, as <fcntl.h> doesn't define it.
> >
>
> What do you mean with some platforms? Some windows platforms?
No, some UNIX platforms (perhaps *all* of them).
> The MSDN states, that fcntl.h will include O_BINARY.
What MSDN states doesn't govern what UN*X platforms do.
> The comment :
>
> /* Win32 needs the O_BINARY flag for open() */
>
> should be changed to something more appropriate, like:
>
> /* Some platforms needs the O_BINARY flag for open() */
No, if it's to be changed at all, it should be changed to something such
as
/*
* On Win32, the "open()" call needs to be given the O_BINARY
* flag, so that data read from the file doesn't have 0x15
* followed by 0x12 turned into 0x12 and so that 0x12 written
* to the file doesn't get turned into 0x15 0x12.
*
* O_BINARY isn't defined on UN*X, so we have a choice of
* #ifdefs selecting one of two "open()" calls (or one of two
* flag argument values in the "open()" call) or defining
* O_BINARY as 0 on UN*X and using O_BINARY in the call on
* all platforms.
*
* We choose the latter, as it's a bit less messy.
*/
> And please note, that O_BINARY on Win32 is defined as 0x8000, so
> defining it as 0 will probably cause problems.
And lease note that the definition as 0 happens only if O_BINARY isn't
otherwise defined; as long as it's done after the include of <fcntl.h>,
that won't be a problem.
