signals

[utzoo!decvax!harpo!npois!ucbvax!dove at mit-dspg]

From: dove at mit-dspg at mit-mc
We are running 2.8bsd, and have a program (dpy) that displays process
activity repetitively after setting the terminal into LITOUT mode.

Occasionally, the interrupt signal one sends to stop the program fails
to execute the routine to which it attached (which restores the
terminal state and exits).  However, a second interrupt in that
situation will always kill the program as though SIG_DFL was in
effect.  It appears that the kernel is failing to execute the signal
connected routine, but is resetting the signal to SIG_DFL.

Is anyone aware of a race condition or bug that could cause this?  I
have tried experimental signal test progs that do lots of terminal
output and they always work.  The only other thing that dpy does is
phys().