Cron mysteriously stops cronning.

[Ross Oliver]

In article <660 at marob.MASA.COM> daveh at marob.masa.com (Dave Hammond) writes
about cron ceasing to function, even though the process remains.

There are two distinct cron problems in the various versions
of SCO XENIX currently in the field.  The first is that cron
runs jobs when it shouldn't, and the second is that cron seems
to go to sleep forever.

The first problem exists in all versions (86, 286, and 386) prior to
release 2.3.1: if more than about 19 hours (64K seconds) is supposed to
elapse between cron jobs, cron will execute the job at the proper time,
and again about five hours later.  This is caused by the fact that the
number of seconds cron is supposed to sleep exceeds the maximum
size of the short int that is passed to alarm(S), so the elapsed
time gets truncated.  This most often happens when there is only one
cron job that executes once per day.  You can work around this bug by
giving cron something to do (like execute /bin/true) at least every
12 hours.  This problem is fixed in release 2.3.1 and later, and a new
version of cron is available from Tech Support for 286 and 386 2.2.3.

The second problem exists only in 386 versions, and is actually a bug in
the operating system itself.  As I understand it, the parameter passed
to alarm(S) by a 286 process (which cron is) is incorrectly sign-extended
when converted to a 386 int.  Thus, the alarm signal is scheduled
to be sent in 20 years rather than tomorrow morning, and cron
goes to sleep for a very long time.  This is fixed in release 2.3.2.
Like the first problem, you can also fix it by giving cron some
extra things to do so it never has to sleep for more than 12 hours.

Ross Oliver
Technical Support
The Santa Cruz Operation, Inc.