threads branch

[James E Keenan]

On 7/7/12 9:53 AM, James E Keenan wrote:
>I sent you email on June 28
> which included the following:
>
> "When I tested the threads branch on Darwin/PPC, there was a stretch of
> about 8 tests in one file which caused 'make test' to grind to an
> absolute halt. My laptop did not freeze, but 'make test' just silently
> stopped.
>

These poor results on this platform persist.  In the threads branch, 
building on Darwin/PPC with --cc=gcc --link=g++ --ld=g++ (same as I've 
done since 2006), 'make test' (running TEST_JOBS=1) hangs on 
t/pmc/nci.t, starting at around test 37.

When I run the individual tests, I get inconsistent results.  More often 
than not, the individual tests pass -- even though the overall test file 
hangs.  Sometimes the individual tests hang as well.

[parrot] 525 $ ./parrot /Users/jimk/gitwork/parrot/t/pmc/nci_42.pasm
ok 1
ok 2
in callback
user data: 42
external data: 88
^C
[parrot] 526 $ ./parrot /Users/jimk/gitwork/parrot/t/pmc/nci_37.pasm
ok 1
ok 2
in callback
user data: 42
external data: succeeded
done.
[parrot] 527 $ ./parrot /Users/jimk/gitwork/parrot/t/pmc/nci_38.pir
created a callback sub
loaded a function that takes a callback
in callback
user data: 42
external data: succeeded
the callback has run
[parrot] 528 $ ./parrot /Users/jimk/gitwork/parrot/t/pmc/nci_39.pasm
ok 1
ok 2
in callback
user data: 42
external data: 77
done.
[parrot] 529 $ ./parrot /Users/jimk/gitwork/parrot/t/pmc/nci_40.pir
created a callback sub
loaded a function that takes a callback
in callback
user data: 42
external data: 99
the callback has run
[parrot] 530 $ ./parrot /Users/jimk/gitwork/parrot/t/pmc/nci_41.pasm
ok 1
ok 2
in callback
user data: 42
external data: succeeded
done.
[parrot] 531 $ ./parrot /Users/jimk/gitwork/parrot/t/pmc/nci_42.pasm
ok 1
ok 2
in callback
user data: 42
external data: 88
done.
[parrot] 532 $ ./parrot /Users/jimk/gitwork/parrot/t/pmc/nci_43.pir
created a callback sub
loaded a function that takes a callback
in callback
user data: 42
external data: 88

####
One instance of './parrot /Users/jimk/gitwork/parrot/t/pmc/nci_37.pasm' 
completed successfully (above), but 'make test' has been hung at a 
previous instance of that test for 18 minutes.

Note also that t/pmc/nci_42.pasm hung to the point where I had to kill 
it, but then it subsequently ran correctly.

So I've now observed inconsistent or poor results in t/pmc/nci.t in the 
threads branch on two different platforms.

I don't think it's ready for prime time.

Thank you very much.
Jim Keenan