[nightly] 15-Oct-2004 build of HEAD on OpenBSD/x86 (gcc-3.3.2)

[Donald Bruce Stewart]

simonmar:
> On 16 October 2004 06:16, Donald Bruce Stewart wrote:
> 
> > dons:
> >> Nightly build started on manzano at Fri Oct 15 23:01:42 EST 2004.
> > 
> >>     3732 expected passes
> >>       34 expected failures
> >>        1 unexpected passes
> >>       20 unexpected failures
> >> 
> >> Unexpected passes:
> >>    galois_raytrace(optasm)
> >> 
> >> Unexpected failures:
> >>    tcfail021(normal)
> >>    tcfail103(normal)
> >>    tcfail104(normal)
> >>    TH_repGuard(normal)
> >>    conc018(threaded)
> >>    conc020(threaded)
> >>    conc023(threaded)
> >>    conc026(threaded)
> >>    conc036(threaded)
> >>    conc038(threaded)
> >         ^^^
> >         These conc0xx tests deadlocked and had to be killed manually.
> >         Something happen yesterday?
> 
> Yes, I made some changes to the threaded RTS.  I don't have any problems
> here, though.  Hmmm.
> 
> Could you run one of them with +RTS -Ds (compile with -debug, you'll
> probably have to compile up a new version of the RTS: add thr_debug to
> GhcRTSWays in build.mk).

manzano$ gmake TEST=conc026 EXTRA_HC_OPTS=-debug WAY=threaded
manzano$ ./conc026 +RTS -Ds 2> /tmp/f

The log is at 
        http://www.cse.unsw.edu.au/~dons/conc026-Ds.txt

It looks like the code fairly quickly drops into a pattern of:
 
        sched (task 0x3c02b800): worker: got capability
        sched (task 0x3c02b800): worker (token 3): re-entering RTS
        sched (task 0x3c02b800): thread 3 did a _ccall_gc
        sched (task 0x3c02b800): worker: released capability
        sched (task 0x3c02b800): worker (token 3): leaving RTS
        sched (task 0x3c02b800): worker: returning; workers waiting: 0

-- Don