I have tried to compile the 1.93 sources on my DEC-Alpha running
DEC-UNIX 4.0 (formerly called DEC-OSF) using gcc 184.108.40.206
Compiling went fine (except for the glob-subdirectory, but I just
removed it, since the functions glob and fnmatch are in the standard
When I used octave, I got core-dumps whenever I tried to call a
command that does not exist. The problem was somewhere in the kpathsea
library. So I linked octave 1.93 with the kpathea library of octave
1.1.1. This seemed to work fine.
After using octave 1.93 for a while I found the following problems:
1. Calculations like 222^222 do not give "Inf" but "1.7977e+308".
2. fscanf does not work properly (not as documented and not as it
works with octave 1.1.1
Both fscanf(fnum, "%s") and fscanf(fnum, "%c") read the whole
contents of a file, instead of just a word or just a character.
Any comments, hints, flames,... are appreciated,
Peter Bruhn | Department of Applied Computer Science
| University of Economics and Business Administration
| Augasse 2-6, A-1090 Vienna, Austria
[hidden email] | Phone: +43 1 31336 x5209 fax: x739
: 1. Calculations like 222^222 do not give "Inf" but "1.7977e+308".
IEEE floating point math is not fully supported by gcc 2.7.2 on the
DEC Alpha. I believe this problem will be fixed with gcc 2.8.0 (the
next major release). No, I don't know any better than anyone else
when that might be available.
: 2. fscanf does not work properly (not as documented and not as it
: works with octave 1.1.1
: Both fscanf(fnum, "%s") and fscanf(fnum, "%c") read the whole
: contents of a file, instead of just a word or just a character.
Sorry, the documentation is lagging a bit behind.
The C-style I/O functions are now compatible with Matlab by default
(it seemed to be what people wanted).