Jason Thorpe a écrit :

> Yes, it's really an error.  IMO, the thread library should abort on  
> such errors.  It indicates that the app modified a data structure it  
> thought was locked, but in fact was not.

Okay, I withdraw my proposal! ;)
On the other hand, though I've not carried out a deep analysis, I wonder 
how a program that is known to work on Linux "as is" can generate this 
kind of error on NetBSD. Aren't both pthreads libraries (functionally) 
identical?

Vincent