Re: test-lock failure on NetBSD 10

[Bruno Haible]

Collin Funk wrote:
> > * tell how many threads your machine supports?
> 
> I created the virtual machine with 8 cores and 8 GB memory.

Ah, this explains why I never saw this: I create most of my VMs with 1 CPU
(so that I can run a dozen of them in parallel).

> >   test-pthread-mutex
> 
> $ time ./gltests/test-pthread-mutex
> Starting test_pthread_mutex_normal ... OK
> Starting test_pthread_mutex_recursive ... OK
> 
> real  0m0.873s
> user  0m1.282s
> sys   0m4.278s
> 
> >   test-pthread-rwlock
> 
> $ time ./gltests/test-pthread-rwlock
> Starting test_rwlock ... OK
> 
> real  0m6.091s
> user  0m0.423s
> sys   0m7.351s

This looks fine. Nothing wrong here.

> >   test-lock
> 
> $ time ./gltests/test-lock          
> Starting test_lock ... OK
> Starting test_rwlock ... OK
> Starting test_recursive_lock ... OK
> Starting test_once ...Alarm clock
> 
> real  10m3.724s
> user  70m3.914s
> sys   9m39.292s
> 
> Here is the output of sysstat(1) from another ssh session:
> 
>                     /0   /1   /2   /3   /4   /5   /6   /7   /8   /9   /10
>      Load Average    11.0|||||||||||||||||||||||||||||||||||||||||||||
> 
>                          /0   /10  /20  /30  /40  /50  /60  /70  /80  /90  
> /100
>   collin 23352 test-lock XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX

The first step towards investigating this one is to compare with the
results of (in a testdir of modules lock, pthread_once, call-once):

  $ time ./test-pthread-once1
  $ time ./test-pthread-once2
  $ time ./test-call_once1
  $ time ./test-call_once2

Then, you might see whether setting ENABLE_DEBUGGING to 1 in the test's
code reveals a pattern...

Bruno