Sat 22 Sep 2012 12:48:52 AM UTC, comment #2:
I found the problem/feature.
When combining a double() with a uint8(), the uint8() type has precedence. Thus, class ([uint8(1), double(1024)]) is converted to uint8, whose largest number is 255.
Ive pushed a changeset
http://hg.savannah.gnu.org/hgweb/octave/rev/283cada76dd6
|
Fri 21 Sep 2012 08:56:38 PM UTC, comment #1:
This bug is only present when using the gnuplot backend, and is present in the developers sources.
Using the script below, I produced the attached plot stream to gnuplot (34720.gp).
I don't see anything wrong with the ascii content, and suspect the problem is hidden in the binary part. After experimenting, I found that converting "y" to a double gives the desired result. So, the following works for me.
It doesn't fix the underlying bug, but a superficial fix (below) is to convert the {x,y,z}data to doubles.
(file #26613)
|
Fri 21 Sep 2012 08:07:12 PM UTC, original submission:
Hello,
in all octave versions > 3.6 I have found a severe plot bug:
A= imread('untersuche.tif'); % chess board
y=A(1, 1:537);
plot(1:537, y)
plot in octave end up at column 265.
plot-command in all Matlab versions > 2007b is ok.
|