Re: TAB*, NL*, BS* etc in <sys/ioctls.h> in GNU/Hurd and GNU/BSD

[Roland McGrath <roland at redhat dot com>]

> Erf, defining OLCUC to 0x10 was actually a bad idea since it's OCRNL in
> BSD's oflags.  In the patch below I used bit 7 since in BSD ioctl that
> was EVENP which is !PARODD cflag in termio. But maybe BSD people will
> define a flag for bit 7 someday so we should choose something else?

OLCUC is a c_oflag bit.  It only matters that it not conflict with other
c_oflag bits.  Since 0x80 is the next one BSD would add for something else,
it might be better to use a higher bit.

> There's also a problem with BSD's CDTRCTS, which is CRTS_IFLOW in GNU
> (part of CRTSCTS).  I don't see how to fix it without either not support
> DTR/CTS on GNU or make existing binaries that use RTSCTS suddenly try to
> use DTR/CTS too (or just half of it).

The Hurd term server doesn't actually support any of those bits, so there
is no practical compatibility problem with changing them for GNU.  Neither
BSD nor Linux supports separate CRTS_IFLOW and CCTS_OFLOW bits, just a
single CRTSCTS bit.  So I think it is fine to just harmonize completely
with the BSD bits here.

> +# ifdef BS1
> +#  undef 
> +# endif

Need VT[01] in these lists too.

> +# define ONOCR	(1 << 5)	/* Discard CR's when on colpumn 0.  */

column

> +#endif /* __USE_MISC || __USE_XOPEN */
> +#define VTDLY	(1 << 20)	/* VT delay.  */
> +#define VT0	(0 << 20)	/* VT delay type 0.  */
> +#define VT1	(1 << 20)	/* VT delay type 1.  */

These belong inside that #if.

> +#ifdef	__USE_BSD
> +#define B7200	7200
> +#define B14400	14400
> +#define B28800	28800
> +#define B76800	76800
> +#endif
>  #define	B57600	57600

There's no need to protect these.  B[0-9]+ are all allowed.
Make the whitespace line up with the adjacent lines.


Thanks,
Roland