[178] in Pthreads mailing list archive
re: errno & multiprocessors (was Re: An idea for pthreads )
daemon@ATHENA.MIT.EDU (Sean Levy)
Fri Oct 27 14:11:48 1995
From: Sean Levy <snl@PANEER.NDIM.EDRC.CMU.EDU>
To: pthreads@MIT.EDU
In-Reply-To: <199510262323.AAA21049@eilis.apocalypse.org>
Reply-To: snl@PANEER.NDIM.EDRC.CMU.EDU
Date: Fri, 27 Oct 95 13:33:41 EDT
One point I haven't seen made for having ports of pthreads for things
like OSF/1 is that it is a very big help to debug using a user-space
package, even if you also have kernel threads. In particular, OSF/1
pthreads as of 2.0 has some really annoying, serious bugs and
deficiencies, and the little help that DEC gives you in debugging is
pathetic. Supposedly, this is much better under 3.2, but I can't use
3.2 for other reasons at the moment. I initially started using MIT
pthreads on OSF/1 for exactly this reason -- to help me debug. This
is why, for instance, the old Mach cthreads API has three different
implementations: a totally user-space one, one based on Mach kernel
threads, and one based on "fake" threads using separate Unix processes
[well, Mach tasks].
--S
--
Sean Levy, n-dim Group, Engineering Design Research Center, Carnegie Mellon U.
snl@paneer.ndim.edrc.cmu.edu 5000 Forbes Ave., PGH PA 15213, USA
He who would be the greatest | http://www.cs.cmu.edu/~snl
must be the servant of all | +1 412 268 5215 {Fax:5229}
