Updates (should appear in issue 8 of the std) have been approved,
as resolutions to bug reports 1053 (stty) and 1151 (the rest),
which add (in ways compatible with what we already have):
SIGWINCH
struct winsize
stty size, stty rows N, stty cols N
Since struct winsize is manipulated via ioctl, and ioctl is not POSIX,
it also adds 2 functions that we don't have
tcgetwinsize()
tcsetwinsize()
Theses were originally to be called tcgetsize() and tcsetsize() which
several other systems already provide, until it was discovered that there
are systems (well, one known) that have functions with those names but
with a different signature (what the difference is, I have no idea, nor
does it matter.)
SIGWINCH goes in <signal.h> where we have it (from <sys/signal.h> like
all the others). The stty changes are no different than what we have
(it is permitted to include the size info in -g output, but not required,
applications are supposed to save/restore the size using
ttysize=$(printf "rows %d cols %d" $(stty size)) # to
save
stty $ttysize # to restore
(obviously other equiv formulations are also OK.)
The rest of the new stuff belongs to <termios.h>
Actions that will be needed as I see it are
1. Move struct winsize from <sys/ttycom.h> to <termios.h>
which probably means moving it to a new <sys/xxxx.h> and
including it from both <termios.h> (outside the _NETBSD_SOURCE
section which already includes <sys/ttycom.h>) and from
<sys/ttycom.h> for backward compat (incl to avoid needing to
adjust other kernel sources.)
2. Add the 2 new functions to src/lib/libc/termios and their
prototypes to <termios.h>
3. Change the man page stty(1) to make "cols" appear as the
actual arg, and "columns" the alias .... (highly crucial that one!)
The proposed "struct winsize" is only required to contain ws_row and ws_col
but they are defined so that other fields (like ws_xpixel and ws_ypixel,
which everyone has, can remain).
The signatures for the new functions (for <termios.h>) are:
int tcgetwinsize(int, struct winsize *);
int tcsetwinsize(int, const struct winsize *);
They do the obvious (simply perform the ioctl, the int is "fd")
They both return 0 (OK) or -1 (error, with errno set) - ie: exactly
what the ioctl returns (they could almost be defined as macros...
but not, as then we'd have to expose the ioctl names to non NETBSD_SOURCE
code.)
The final text for the 2 resolutions is in:
note 3863 of: http://austingroupbugs.net/view.php?id=1053 (stty)
note 3856 of: http://austingroupbugs.net/view.php?id=1151 (the rest)
kre
ps: Issue 8 is still some time away - definitely not this year, quite
likely not next... so this is not urgent, but getting it done to include
in NetBSD 8 would not hurt, Issue8 is likely to appear during NetBSD 8's
lifetime.
