Re: [PATCH 00/16] sys/ioctl.h and termios.h cleanups

[Rich Felker <dalias@libc.org>]

On Sun, Jul 03, 2016 at 05:01:24PM +0200, Szabolcs Nagy wrote:
> the contents of sys/ioctl.h is linux specific and not defined by any
> standard, the defacto standard is whatever glibc sys/ioctl.h does.
> 
> termios.h is defined by posix, but there are various extensions.
> glibc does not include linux uapi headers, but contains a copy of
> most asm/termbits.h definitions etc.
> 
> this is v2 of the patchset, changes:
> - add powerpc64 and mipsn32 changes
> - add *speed members to mips termios struct
> 
> Szabolcs Nagy (16):
>   use the generic ioctl.h for x86_64, x32 and aarch64
>   remove linux asm/sockios.h macros from ioctl.h
>   remove ioctl macros that were removed from linux uapi
>   add missing SIOCSIFNAME from linux/sockios.h to ioctl.h
>   add missing TIOC* macros to ioctl.h
>   remove mips and powerpc ioctls that are missing from linux uapi
>   fix mips, mips64, mipsn32 TIOCM_* macros in ioctl.h
>   fix TIOCMSET in mips ioctl.h
>   remove termios2 related ioctls from sh ioctl.h
>   fix FIOQSIZE in arm ioctl.h
>   fix generic termios.h to be more consistent with glibc
>   fix powerpc termios.h to be more consistent with glibc
>   fix mips termios.h to be more consistent with glibc
>   fix TCS* definitions in mips termios.h
>   fix CBAUDEX in powerpc termios.h
>   fix struct termios in mips64 and mipsn32 termios.h

Thanks! I'm merging all of these except #2, which broke the build of
musl itself by eliminating SIOCATMARK, and #16, for which I already
had my own version with a much more detailed commit message discussing
the motivation.

#11-13 also break musl itself (CBAUD needed) but I'm introducing an
extra commit before them to use the _BSD_SOURCE FTM. I also renamed
them (as discussed on irc) to make it clear that the goal is namespace
correctness and not just matching glibc (which would only be correct
if glibc is more right, which it is in this case).

#2 could still be considered at some point, but if we do that we need
to find a way to expose to musl the arch-specific SIOCATMARK it needs
for internal use.

Rich