On Feb 20, 2007, at 5:42 PM, Bucky Katz wrote:

> You seem to misunderstand what's being asked for.  Perhaps you're not
> aware of what's going on in the evbarm port just now?

I poke my head in and look at it once in a while.  Sure, it's broken  
right now.  BFD -- this is -current, and sometimes things are broken.   
It's not like this is the first time that's happened (and certainly  
not with just the evbarm port).

> Nobody changed the subject line, but the topic now is a) fix omap in
> the face of newlock2 and b) put m:n back in as an option for
> uniprocessors.
>
> There's no one arguing for "going back".
>
> There's no one asking for work to be thrown away.

I think YOU misunderstand what you're asking for.  Here's the simple  
truth: Fixing M:N in the presence of newlock2 involves a lot of work.   
Either:

1- The SA code needs to be almost entirely rewritten, even for  
uniprocessors.
-or-
2- Someone needs to write a new threads library that implements M:N  
without SA.

Neither of those are going to be accomplished in the short term.  So,  
to achieve your goal in the short term, newlock2 would need to be  
backed out (which is the original request as indicated by the Subject:  
on this thread).  This would in fact mean "throwing away" work that,  
by all of the information that I have, the Project as a whole  
considers valuable and is unwilling to do.  I'm not saying that as the  
All Powerful Dictator Of The Project (ha!  If only that were true,  
muahahaha!), but rather as an observer of the Project who happens to  
be friends with many people in the Project's leadership and has been  
pretty deeply involved in these issues in the past (and to a lesser  
extent in the present).

> B is what I _thought_ I was trying to work out with core before you
> issued your fiat, which, apparently, you're in no position to issue,
> as, as you say, you *don't* speak for core.

Well, it's certainly your right to ignore me entirely.

> Perhaps it would be better if the people who came late to the table
> sit quietly and let those who are trying to find a win/win do their
> thing?

Look at it from my perspective -- I've been involved with the newlock2  
stuff since 2002, and while Nathan Williams was the author of the SA- 
based pthreads implementation, I was the one who stayed up all night  
to merge the branch to the trunk.  In fact, I was also the person who  
wrote the ARM support for SA and pthreads (from that perspective, I  
was somewhat disappointed to see SA-based-M:N disappear, but I  
understand that losing it now helps further the Project's longer-term  
goals).  So, to me, it is YOU who are the person who has come late to  
the table.

-- thorpej