too many SIGALRM's in sc 6.1 (on SysVr2)
Greg Woods
woods at tmsoft.uucp
Wed Apr 5 09:00:24 AEST 1989
I've been using SC quite successfully for some time (since v4.x) on
various flavours of Unix machines (including at least one of Xenix
III, Xenix V, BSD4.1, SunOS-3, SysVr2, and SysVr3) and have been quite
happy with it.
However, today, well into creating my new monthly budget, sc simply
exited, with my shell (csh) telling me "sc: Alarm clock". Of course
nothing was saved, not even a core-dump.
Perhaps the problem stems from the curses library's use of alarm() in
getch(). (My problem occured on and NCR Tower 32/600 running
R01.03.02, and I've compiled sc as per SysVr2 and terminfo/curses.)
The only other possibility, which *should* be impossible in any case,
is that setting SIGALRM to time_out(), which would longjmp(wakeup)
before the jmp_buf is initialized, has caused problems. Perhaps
SIGALRM should be set to SIG_IGN for terminfo/curses compatability.
One other possiblity (that I hope and pray is not true) is that the
kernel let a stray SIGALRM trough from a background process to the
rest of the process group (I run monthd). This is too horrible to
think about, and highly unlikely!
I would appreciate hearing any suggestions, or such.
--
Greg A. Woods.
woods@{{tmsoft,utgpu,gate,ontmoh}.UUCP,utorgpu.BITNET,gpu.utcs.Toronto.EDU}
1-416-443-1734 [h] 1-416-595-5425 [w] Toronto, Ontario, Canada
More information about the Comp.sources.bugs
mailing list