Re: zillion DO problems again

[Adam Fedor]

On Sat, 2003-08-30 at 04:33, Stefan Böhringer wrote:
> I resend this, since I didn't see the original message come through.
> ------------------
> Hi Richard,
> 
> after upgrading to gcc 3.3/cvs as of beginning of week several things
> went better. E.g. the gworkspace finder would no longer crash.
> Unfortunately I still have DO problems. Attached is a version of zillion
> that exhibits the problems but should not differ in the relevant parts
> from versions I sent earlier.
> The problem is now that DO method calls take very long time at some
> points to come through.
> 
> If you use the usual setting:
> make debug=yes; Terminal 1: ./shared_debug_obj/zillion; Terminal 2:
> ./shared_debug_obj/zillionClient; Terminal 3:
> ./shared_debug_obj/zillionLoader simpleZillionJob.bundle
> 
> the client (terminal 2) will hang after printing:
> 2003-08-21 14:23:27.355 zillionClient[23001] Ticks spent executing job 4
> 2003-08-21 14:23:27.372 zillionClient[23001] 80fd788
> 
> This is in -jobUnitDidEndForJobId:seed:(ZillionClient) just before a DO
> call to decreaseLoad. After some minutes (say 10) the line
> 
> 2003-08-21 14:33:27.387 zillionClient[22996] decrease load being called
> 
> is being printed, indicating that the method was indeed called after
> that delay. From that point on all DO methods seem to take that amount
> of time to come through. Can you reproduce this bug?


I can reproduce it (on ppc, gcc 3.2.2) but I don't know why it happens.
It's exactly 10 minutes for me (600 seconds). I wonder if that means
something...