Re: Nested sit-for's

[Richard Stallman]

    In the case where (load-average) is too high, the idle timer may not
    get control back until up to 30 seconds later:

We certainly don't want this timer function, as currently written, to
run from within another timer.

I just tried writing the code to eliminate this loop and make the
timer reschedule itself.  The only hard part was that there is no way
for an idle timer to arrange to run something after a few seconds more
of idleness.  To enable that, I added a function current-idle-time.
With that, it is easy to write the code for jit-lock to avoid looping.

Which means we can simply say that timer functions should not call
sit-for.

I think some timer functions will still need to call
accept-process-output, but we can say they are supposed to call it in
the way that doesn't run timers.

Does anyone think that will cause a problem?