cvs commit: src/lib/libpthread Makefile src/lib/libpthread/test sigsuspend_d.c src/lib/libpthread/thread thr_cancel.c thr_concurrency.c thr_create.c thr_find_thread.c thr_init.c thr_kern.c thr_nanosleep.c thr_private.h thr_sig.c

[David Xu]

----- Original Message ----- 
From: "Daniel Eischen" <eischen at vigrid.com>
To: "Scott Long" <scottl at FreeBSD.org>
Cc: "David Xu" <davidxu at FreeBSD.org>; <src-committers at FreeBSD.org>; <cvs-src at FreeBSD.org>; <cvs-all at FreeBSD.org>
Sent: Saturday, July 19, 2003 5:10 AM
Subject: Re: cvs commit: src/lib/libpthread Makefile src/lib/libpthread/test sigsuspend_d.c src/lib/libpthread/thread thr_cancel.c thr_concurrency.c thr_create.c thr_find_thread.c thr_init.c thr_kern.c thr_nanosleep.c thr_private.h thr_sig.c thr_sigmask.c ...


> On Fri, 18 Jul 2003, Scott Long wrote:
> 
> > David and all,
> > 
> > Thanks for all of the good work here!
> > 
> > There has been a grwoing question of which threading package will be
> > made 'default' for 5.x.  The main requirement is that the default
> > threading package be supported on all Tier-1 platforms.  Right now,
> > only libc_r meets that requirement.  Here's a quick table from
> > memory, please correct me if I'm wrong:
> > 
> >          i386  alpha  sparc64  ia64
> > c_r       X     X       X       X
> > kse       X                     X
> > thr       X             X       X
> > 
> > Don't forget that amd64 is up-and-coming also.
> 
> I have most of the amd64 userland bits.  The syscalls
> *context() don't seem to work correctly yet (testing
> done on sledge).  I can commit them so that they are
> there for others to play with.
> 

Can you just commit it ? this let any parties interested
in KSE can work on it.

> As Marcel can probably confirm, it is not that difficult
> for someone with <arch> knowledge to write the necessary
> libkse bits -- probably no more than a weekend.  Marcel's
> ia64 bits are a good path to follow.
> 
> I'd like to see the emphasis put more on the <arch>s; if
> they are going to be Tier-1, provide the necessary libkse
> bits :-)
> 
> -- 
> Dan Eischen