[chuck-users] time goes backwards (and I don't like that)

[Kassen]

Rob;

*Whew* -- thanks -- I was worried that I was suffering a dissociative fugue
> (not to be confused with a counter-fugue).
>

Just because ChucK has bugs doesn't mean you're not insane, these two
factors might be independent :¬p.


>
> A hint to the hunting party: the "wrong" time returned by now always seems
> to be an integer.  It's as if the fractional part is getting lopped off -
> either by truncation or rounding, I'm not sure which.
>

There is definitely some rounding error going on. Here is a edit of the code
that lops off some zeros. This version never prints "bang" in 1.2.1.2.

=================
Event _e;
time _t;
time foo;

spork ~ beatProcess();

while (true) {
   now + 1.01::second => _t;
   _e => now;
   if (foo > now) <<<"bang">>>;
   now => foo;

   <<< "returned from _e => now at", now >>>;
}

fun void beatProcess() {
   while (true) {
       _t => now;
       if (foo > now) <<<"bang">>>;
       now => foo;
       <<< "broadcasting at", now >>>;
       _e.broadcast();
       //me.yield();
       samp => now;
   }
}
==========================



>
> Fetch me my halberd!
>

I think it's hiding in rounding in the Shreduler, let's set fire to the
shreduler and see if it comes out!

In all seriousness; this sounds tricky, I reported a hard to pinpoint
execution order error to Ge before and said I suspected something had
changed in the shreduler since last version. Ge replied he couldn't remember
changing anything to it. The odd thing is that the bug I saw would come and
go with using different samples (as in .wav files) but clearly samples only
come in integer durations if we express them in samp's.

Grumble. Oh, well, at least we've got it cornered.

Kas.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/chuck-users/attachments/20090314/7e94ea21/attachment.htm>