Marcus, a suggestion to try....
On Tue, Oct 24, 2000 at 05:43:23PM +0200, Marcus Brinkmann wrote:
>
> Yes, also strange, because the code is almost the same.
> Also, opening/closing the kbd device 5000 times in a small program also
> worked fine.
Well, the translator code is almost the same, but the X code is not.
The only differences I note in the translators seem insignificant or
irrelevant. They are:
different buffer size (do size = 1 and != 1 require different
handling? this seems like a boundary case)
mouse has option support
kbd has ioctl support
slight difference wrt demuxer
slight difference in close_hook
slight difference in *_thread ncopy loop (is one off-by-one?)
In X, the differences seem more significant. Ref:
.../Xserver/hw/xfree86/os-support/hurd/hurd_init.c:xf86OpenConsole
.../Xserver/hw/xfree86/os-support/shared/posix_tty.c:xf86OpenSerial
The mouse open in xf86OpenSerial looks like any ordinary fd open, except
it's wrapped with SYSCALL which is just:
#define SYSCALL(call) while(((call) == -1) && (errno == EINTR))
The keyboard open in xf86OpenConsole however uses get_privileged_ports()
and mach_port_deallocate(), i.e., it is mach/hurd-specific. I don't
known what that means, but they're not the same. Just hacking around
to see what would happen, I did a few variations: taking out the
get_privileged_ports() and mach_port_deallocate() calls to make the kbd
code look like the mouse code resulted in "computer bought the farm"
errors. Putting them back and just leaving in a SYSCALL wrapper as is
done for the mouse code results in a hang. That's about when I decided
to go looking at the translator code for differences. Obviously, I'm
just fishing here....
The next thing would have been to explore the differences in the X code
some more (since your kbd test program seems to like the translator code).
It occurs to me that since you already have a working kbd test program,
to try patching it into xf86OpenConsole and see what happens.
Steve
--
Steve Bowman <sbowman@frostwork.net> (preferred)
Buckeye, AZ <sbowman@goodnet.com> <bowmanc@acm.org>
<http://www.goodnet.com/~sbowman/>
Powered by Debian GNU/Linux and GNU/Hurd <http://www.debian.org>