Re: FYI, (brand new?) test-lock failure on Darwin, via nixos/hydra

[Bruno Haible]

Jim Meyering wrote:
> At least on darwin, using the very latest from gnulib,
> 
>   http://hydra.nixos.org/build/1731154
> 
> that shows how grep's gnulib-tests exposed one failure:
> 
> building test-lock.log
> FAIL: test-lock

And the test's log [1] says:

  FAIL: test-lock (exit: 134)
  ===========================

  Starting test_lock ... OK
  Starting test_rwlock ...

Since on this platform the glthread/lock's gl_rwlock_t is the same as
pthread_rwlock_t, it means that one of the functions
  pthread_rwlock_init
  pthread_rwlock_rdlock
  pthread_rwlock_wrlock
  pthread_rwlock_unlock
  pthread_rwlock_destroy
must have failed. Possibly because the machine was too loaded.

I don't know what a program could reasonably do when it wants to
acquire a lock and the OS function to do so fails. Therefore the
test simply aborts in this case.

Bruno

[1] 
http://hydra.nixos.org/build/1731154/download/1/nix-build-lhwwksswzz529m3886kh5hppjwwp93dw-grep-2.10.57-0dc6.drv-0/grep-2.10.57-0dc6/gnulib-tests/test-lock.log