From octave-graphics-request at bevo dot che dot wisc dot edu Mon Sep 17 09:54:48 2001
Subject: Re: new version of TK octave
From: Paul Kienzle <pkienzle at users dot sourceforge dot net>
To: =?iso-8859-1?Q?Jo=E3o?= Cardoso <jcardoso at inescn dot pt>
CC: Paul Kienzle <pkienzle at users dot sourceforge dot net>,        octave-graphics@bevo.che.wisc.edu
Date: Mon, 17 Sep 2001 10:51:40 -0400

Octave is certainly NOT thread safe.  You need to link it against pthreads so that
the threadsafe routines in glibc are picked up correctly.  If not, then the connection
to X fails unexpectedly.  This is also true for Rafael Laboissiere's gtk widgets,
though the symptoms are a little different.  See
http://www.octave.org/mailing-lists/octave-graphics/2000/29

The thread-based implementation of tk_octave still has concurrency
issues.   Particularly, access to octave variables is unprotected.  The easiest
solution is to block all access to octave variables while the interpreter is
running, but I haven't done this yet.  At present users and GUI coders
have to be careful.  I've tried to generate an error by changing a value
in the interpreter while it is being accessed by the GUI but I wasn't able to.

Note that Rafael's gtk widgets process GUI callbacks in octave even when
octave is running something else!  I wasn't able to generate an error there
either, but I didn't try very hard.

Paul Kienzle
pkienzle at users dot sf dot net

Joćo Cardoso wrote:

> In
> http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/octave/tk_octave/readme.txt?rev=1.1&content-type=text/vnd.viewcvs-markup
>
> You say:
>
> Compiling Octave with tk_interp.cc
> ==================================
>
> In order to use tk_interp, we needed to relink the Octave binary with
> pthreads.
>
> And Octave is already thread safe?
>
> Joao