Re: Please test new NSLock implementation!

[Philippe Roussel]

Le mercredi 02 septembre 2009 à 15:55 +0100, David Chisnall a écrit :
> Hi Philippe,
> 
> Thanks for the report.  It seems that the default behaviour of a  
> pthread mutex is subtly different between FreeBSD and with glibc.   
> Mutexes with no explicit attributes are error-checking on FreeBSD /  
> Darwin, but not on Linux.  Given that the test is failing there, and  
> not earlier, it seems that error-checking mutexes do work correctly if  
> explicitly requested.  Hopefully this is fixed in r28599 - please let  
> me know if that works for you.
>
> There are two expected failures for that test.  OS X doesn't throw an  
> exception in double-lock cases, it just logs a warning and deadlocks.   
> The new code logs the warning but doesn't deadlock.  Possibly entering  
> a deadlock state is better, because it forces people to fix their  
> code, but I've left it as-is for now.

Hi David

It no longer locks but emits 2 FAILs

FAIL: Recursive lock with NSLock raises an exception
FAIL: Recursive lock with NSConditionLock raises an exception

Is that what you expexted ?

Philippe