On Sat, Oct 13, 2007 at 03:41:49PM -0700, Alan Irwin wrote:
> On 2007-10-12 21:52+0100 Andrew Ross wrote:
> >As you may have noticed, I've commited some changes to the octave
> >bindings in svn. As a result it is now possible to build plplot using
> >octave 2.9.
> >The basic plplot bindings work fine (as far as I can tell). All the
> >x??.m scripts produce identical results to the C equivalents.
> I have tested on the Debian oldstable (sarge) version of octave (i.e.,
> octave 2.1.69). The build, build-tree test with ctest, install, and
> install-tree test with plplot-test.sh completed without obvious problems.
Good to hear.
> The C and Octave results are identical on my system with two exceptions.
> Thanks, Andrew, for doing this extensive work to get the latest octave 2.9
> working properly (I hope Orion confirms that for his Fedora platform)
> while obviously retaining a good result for octave 2.1.69 for my debian
> oldstable platform.
> The two exceptions mentioned above are for x20 and x21.
> The x20 C/Octave difference seems due to a visible but relatively small
> change in contrast for the "Lena" image when viewed by gv. I have no idea
> what could be causing that. Andrew, if you cannot confirm that contrast
> change for octave 2.9, I wouldn't worry about it.
Alan, you are correct. There was a problem in the code for reading pgm
file in Octave. I don't know how long this has been there. A CR-LF was
not properly handled (perhaps the EOL status of lena.pgm has changed?)
Anyway, it is now fixed.
> The x21 C/Octave difference is caused by different pseudo-random data for
> the two languages and also by different timings (which cause different
> labels) from one run to the next. If somebody is game for this, it would be
> good to change x21c.c (and all other implementations of example 21 including
> the octave one) to a form that gives absolutely consistent results. That
> means replacing the current inconsistent random x/y values by scattered but
> consistent values obtained in some other way and labelling the plots with
> something different than execution time.
I agree. This is a good example, but the non-deterministic nature of it
makes it irritating for testing purposes.
As a further addition I've now added plptex3 / plmtex3 support to the
octave bindings and implemented example 28.