[196] in Pthreads mailing list archive
Re: pthreads
daemon@ATHENA.MIT.EDU (David Brownell)
Thu Nov 2 10:10:30 1995
Date: Thu, 2 Nov 1995 06:25:23 -0800
To: "William S. Gribble" <grib@cs.utexas.edu>
From: David Brownell <brownell@ix.netcom.com>
Cc: pthreads@MIT.EDU
>Making GC work with multiple thread stacks shouldn't be all that hard,
>just a little hacking. The important issue is how much do you want to
>put Guile in bed with threads -- a configure-time option, or do you
>distribute MIT pthreads (GPL'ed already, and builds on SunOS/Solaris/
>Linux/NetBSD/HPUX and I think Irix) with guile and then assume that you
>have the POSIX API already? Hm.
The POSIX API set doesn't include calls like "get all the thread stacks"
and other stuff you need to write a portable garbage collector. Until
there's agreement on what those APIs need to look like, this kind of
functionality is very dependant on what implementation of pthreads you
are using. A full suite of "debugger APIs" would suffice, but of course
it's overkill since you don't need to do stuff like set breakpoints or
start/stop individual threads order to GC an address space.
So on platforms such as Solaris 2.5 that come complete with POSIX.1c
support, life gets a bit more interesting!
- Dave
