Sinabro

So, I tracked down the problem with plotting. It turns out it’s my configuration file, or rather, more precisely, one of my custom functions.

The problem persisted even when I compiled GNU Octave by hand (and the gnuplot, to be sure), but when I removed .octaverc, everything worked fine.

Working backward from that fact, all I know is that some script in my ~/octave must be clobbering a system function, possibly used by plot command.

I suppose the most mysterious thing is … I have another computer that has nearly identical setup (even the same ~/octave!) and everything works fine there, too! Figures …

I think I’ll be switching to Python for scientific computing anyways. I think I finally found a way to do least-squares fit in Python (http://linuxgazette.net/115/andreasen.html) relatively quickly (no, the version in ScientificPython is way too slow, compared to, e.g. GNU Octave). In a way, the main limitation I see for GNU Octave is its MATLAB compatibility, because MATLAB compatibility means both the lack of object-oriented programming and its brain-dead syntax.

Damn. I’m doing it again.

That is, I’m building Octave (v3.0) myself. For mysterious reasons, octave3.0 in Debian testing (or is it in unstable?) doesn’t work with gnuplot for plotting. I mean, if I use __gnuplot_plot__ with raw gnuplot commands, it seems to work fine, but I’m not using deprecated commands in my scripts.

Hopefully this time, I won’t run into problem building the octave-forge.

Sigh. I am evermore waiting for lenny to get released. Hopefully when lenny gets released (and, no, I am not dist-upgrading to lenny) I won’t have to compile my own custom kernel, and hopefully I can go back to uswsusp.