From: rajat@watson.ibm.com (Rajat Datta) Subject: Re: Suspending emacs and bringing it back Date: 4 May 1992 19:59:11 GMT
In article <8a66JB3w165w@csource.oz.au> david@csource.oz.au writes:
>
>It might pay to wait for 0.96 to see if the problem persists.
>
>
Yup, that's what I thought. I'm waiting.
However, I did write a simple program that simply puts the tty into
raw mode and accepts input in an infinite loop. It also enables
SIGCONT, so that I can put it to sleep and when I bring it back, it
can print out the current settings of the tty. This points out a bug
in the tty driver.
Basically, to put a tty into raw mode, you have to turn of CANONical
processing. I also set the VMIN (minimum number of chars input before
responding) to 5 and VTIME (length of time to wait in 1/100 of a
second) to 10. When I receive a SIGCONT, I print out the current tty
settings and go back to reading the input. What I'm finding is that
once I come back from SIGCONT, the tty settings claim to still have
CANONical processing turned off, but the driver is still doing such
processing. That is, I'm getting echoing and cooked mode processing.
I haven't read the code all that closely yet (haven't had all that
much free time lately), but I remember there is a current terminal tty
structure and the tty structure for each process group. Looks like
maybe they aren't set the same in certain occasions when they should
be. Anyway, it all sounds suspiciously similar to the TIOCSPGRP bug
and maybe the same fix will fix both problems.
-- rajat (rajat@watson.ibm.com)