On performance of shouldDoDUChainReferenceCounting again

[David Nolden]

Am Mittwoch 13 Januar 2010 00:04:16 schrieb Milian Wolff:
> I'm sadly not very educated in this whole threading/locking issue, but if
>  you got the nerve for one more of my questions:
> 
> If one disables the wakeOne call, what happens to waiting locks? Sure, the
> case might be rare, but it would still lead to a sleeping thread, no?
> 
> Anyways, if you are not sure this works, i.e. if it could easily lead to
> unstable behavior, I'm against pushing the performance here. KDevelop 4.0
> should be stable, not neccessarily blazingly fast (even though it is
>  already imo).

It cannot lead to unstable behavior, in worst case it could lead to suboptimal 
performance in multithreaded parsing.