Mail Archives: cygwin/1999/09/05/21:40:11
Chris,
> On Sun, Sep 05, 1999 at 10:23:29AM +0200, Victor Szel wrote:
> >> >#ifdef __CYGWIN__
> >> >/* TODO: Get Cygwin fixed so that this bug fix won't be needed */
> >> >static inline int FixCygwinIOflags( int flags )
> >> >{
> >> > /* Starting with O_CREAT, the Cygwin I/O flags are 1 bit too high */
> >> > return ( ( flags & 0x1FF00 ) >> 1 ) | ( flags & 0xFF );
> >> >}
> >> >#endif
> >>
> >> System O_CREAT (octal) O_CREAT (hex)
> >> ----------------------------------------------------
> >> Cygwin 01000 0x200
> >> Digital UNIX 01000 0x200
> >> FreeBSD 01000 0x200
> >> Irix 00400 0x100
> >> Linux 00100 0x040
> >> SCO UNIX 3.2 00400 0x100
> >> Solaris 00400 0x100
> >> Ultrix 01000 0x200
> >>
> >> What was the bug again?
>
> If you are using the open in cygwin then it open() will correctly use
> O_CREAT. It would be inconceivable that this could be wrong. If it
> was, then hundreds of applications would not work.
>
> If you're using cygwin headers in a non-cygwin (aka mingw) application,
> then, um, that's wrong. You should be using another set of headers.
> I'm sure that someone will coment on this and let you know which they
> should be.
Yeah really strange.
One more idea: we are linking MOLDNAME lib, couldn't that be a problem.
BTW we are including <mingw32/share.h> and <fcntl.h>, so the flags should
be OK, but it seems that the open() function is executed from the MFC DLL
which uses different flag values.
Victor Szel
--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe AT sourceware DOT cygnus DOT com
- Raw text -