Re: bug #40256

[Rik]

On 11/08/2013 10:00 AM, address@hidden wrote:
>>> Is #40256 (datetick doesn't remove previous X-grid & ticks) still in the
>>> >> list? (it is a regression and actually a nasty one for me)
>> > 11/7/13
>> > 
>> > Philip,
>> > 
>> > The bug is listed on the 3.8 Bug Fix list
>> > (http://wiki.octave.org/3.8_Bug_Fix_List), however it's not listed as an
>> > official gating item for the release.  I took a quick look at it, but I
>> > don't spend much time with gnuplot so I don't have a good answer.  The
>> > problem is not with datetick, but rather with how the command "set (gca,
>> > 'xticklabel', ...)" works.  Instead of just overwriting the existing
>> > labels, this is apparently invoking the auto-axis labeling routine. 
>> > Simple
>> > code to reproduce:
>> > 
>> > plot (1:4)
>> > ## At this point labels are 1, 1.5, 2, 2.5, 3, 3.5, 4
>> > set (gca, 'xtick', 1:4)
>> > ## At this point labels are 1, 2, 3, 4
>> > set (gca, 'xticklabel', {'A', 'B', 'C', 'D'})
>> > ## Now labels are 'A', 1.5, 'B', 2.5, 'C', 3.5, 'D'
> The underlying issue is that because of bug #33748 and bug #40246 I need to
> revert to gnuplot. Now gnuplot also bugs out on me. 
> Rescaling before handing off to fltk/OpenGL is quite cumbersome as we'd need
> to patch all code (many of which is Matlab code still used in Matlab)
> invoking plotting. That's hundreds of functions to process....
>
> Philip
11/8/13

Philip,

I hear you.  The use of floats, rather than doubles, by OpenGL is quite a
problem.  The workaround should be in the graphics engine so that only one
fix needs to be made and everyone else, the hundreds of m-files in your
case, benefits.  But, that isn't going to occur for the 3.8.0 release.  I'm
trying an hg bisect to see if I can find where the xticklabel problem got
introduced to Octave.  Already I know it was before Jan. 1, 2013.  But this
isn't my area of the code base so I'm mostly hoping I can identify the
problem area for someone else to fix.

--Rik