jalsop@seachg.UUCP (John Alsop) (05/03/90)
As I understand things, most Tower applications software should be ignoring
the SIGPWR signal, as all the power recovery cleanup happens behind the
scenes.
It seems that in SVR3 on both the 4x0/6x0 and [357]00 versions, "vi" is not
ignoring this signal. This causes strange behaviour when a power recovery
occurs and you are in a child process of vi.
To reproduce the problem:
$ cd /tmp
$ vi
:!sh # start a subshell
[ Next few lines are interaction with the sub-shell ]
$ ps -f # show process id's
UID PID PPID C STIME TTY TIME COMMAND
jalsop 3922 1 0 08:44:24 b 0:12 -csh
jalsop 6988 3922 0 17:29:12 b 0:04 vi
jalsop 7080 6988 0 17:36:55 b 0:00 sh
[ Cause a SIGPWR either by turning the machine on and off, or sending a
SIGPWR explicitly to the vi process]
$ kill -19 6988
[ At this point, vi wakes up, thinking that the sub-shell has finished.
As a result, both vi and the shell start grabbing input from the tty at
the same time, causing mass confusion. You will most likely have to
turn your terminal off and on to clear the situation.]
This problem does not occur on SVR2 (Release 2.01.01) nor does it occur
on SVR3 with programs that start a sub-shell using the system() call
(without enabling the SIGPWR signal). Therefore I assume that vi has enabled
this signal on purpose.
Can anyone suggest why this is occurring, or possible ways to circumvent
the problem.