I notice QingLong has made some changes today as well, so I hope I'm not
stealing QingLong's thunder here.
In response to frequent user complaints (which I will forward here in a
moment),
I've changed rlimit from an error to a warning, and also eliminated the
slow startup "feature" when queued is not in debug mode.
(Also, what are these comments about "dumbhead" for DEBUG in the
messages header file? :) Yes, of course, we should use DEBUGG as the
symbol for the debug mode. DEBUG conflicts with the old Lotus PC
compiler, amongst other places. If there was any use of the DEBUG
symbol, it was heriditary from the old batch code.)
Please try the new code!
*March 13, 2001 W. G. Krebs <wk...@gn...>
Changed itorl to make invalid integer rlimit a warning (and then
only in debug mode) instead of a fatal error as it was.
I was getting a lot of user complains about "invalid integer
rlimt" causing queued to coredump.
The problem is that some operating systems have different numbers
of rlimits, and it is possible that a number can be invalid on
one system and valid on another. The best solution is simply
to change this to a warning on the affected systems. Rlimit is
by no means a necessary operation, and it should be no means
cause a coredump!
I've also fixed another frequent source of user complaints.
the slow startup "feature" that is supposed to prevent
Queue from flooding the system with jobs when the daemon is
started. in practice, this just annoys users starting queued
in non-debug mode, since they can't understand why they must
wait for the system to startup. This around line 930 in queued.c,
there are additional comments in the code suggesting another
way to implement a slow-startup that would be less annoying to
users but still prevent the system from being flooded by a full
job queue.
|
