Bug in plotting - 3.0.4 RC1

[Jaroslav Hajek]

On Wed, Jan 21, 2009 at 5:12 PM, Ben Abbott <bpabbott at mac.com> wrote:
> On Wednesday, January 21, 2009, at 09:16AM, "Marco Caliari" <marco.caliari at univr.it> wrote:
>>>>>> x = linspace(0,10,11);
>>>>>> y = x+1;
>>>>>> plot(x,y,x,y)
>>>>>> xlabel('x')
>>>>>> ylabel('y')
>>>>>>
>>>>>> produces the following
>>>>>>
>>>>>> error: __go_draw_axes__: unknown object class, root figure
>>>>>> error: evaluating switch command near line 332, column 7
>>>>>> error: evaluating for command near line 328, column 5
>>>>>> error: evaluating if command near line 25, column 3
>>>>>> error: called from `__go_draw_axes__' in file
>>>>>> `/usr/local/share/octave/3.0.4/m/plot/__go_draw_axes__.m'
>>>>>> error: evaluating switch command near line 58, column 4
>>>>>> error: evaluating for command near line 56, column 2
>>>>>> error: evaluating if command near line 33, column 7
>>>>>> error: evaluating if command near line 26, column 5
>>>>>> error: evaluating if command near line 25, column 3
>>>>>> error: called from `__go_draw_figure__' in file
>>>>>> `/usr/local/share/octave/3.0.4/m/plot/__go_draw_figure__.m'
>>>>>> error: evaluating if command near line 81, column 6
>>>>>> error: evaluating if command near line 78, column 4
>>>>>> error: evaluating if command near line 76, column 2
>>>>>> error: evaluating for command near line 75, column 7
>>>>>> error: evaluating if command near line 45, column 5
>>>>>> error: called from `drawnow' in file
>>>>>> `/usr/local/share/octave/3.0.4/m/plot/drawnow.m'
>>>>>>
>>>>>> I'm using Octave-3.0.4 RC1
>>>>>>
>>>>>> No problem with Octave-3.0.3.
>>>>>>
>>>>>> Best regards,
>>>>>>
>>>>>> Marco
>>>>>
>>>>>
>>>>> I fixed the problem changing back
>>>>>
>>>>>    Matrix new_kids (n-1, 1);
>>>>>
>>>>> to
>>>>>
>>>>>   Matrix new_kids (1, n-1);
>>>>>
>>>>> as it was in octave-3.0.3/src/graphics.cc
>>>>
>>>> Not sure it is the right fix. In fact, now there is a problem with the
>>>> order of the legends, which can be fixed changing back
>>>>
>>>>   for i = nargs:-1:1
>>>>
>>>> to
>>>>
>>>>   for i = 1:nargs
>>>>
>>>> as it was in octave-3.0.3/scripts/plot/legend.m
>>>>
>>>> Marco
>>>
>>> Regard this problem, it originated due to children having a reverse order as
>>> compared to the commercial variant. The last change you made should reverse
>>> the order of the legend, which is not correct (1st line at the bottom and
>>> last at the top). Can you confirm?
>>
>>Suppose you have legend('a','b'): WITH the last change I have 'a' at the
>>top and 'b' at the bottom. And, what is more important, the label 'a'
>>correctly refers to the first plot and 'b' to the second. WITHOUT the last
>>change I have 'a' at the bottom and 'b' at the top. But, the label 'a'
>>wrongly refers to the second plot and 'b' to the first.
>>
>>> Regarding your original error, "unknown object class, root figure", can
>>> you/someone confirm exactly where that occurs in __go_draw_axes__? (I don't
>>> have a copy of 3.0.4 available on my Mac).
>>>
>>> Browsing the sources, it looks like it is line 941/942 below
>>
>>Yes.
>>
>>> 314 switch (obj.type)
>>> 315 case "image"
>>> [...]
>>> 940 otherwise
>>> 941 error ("__go_draw_axes__: unknown object class, %s",
>>> 942 obj.type);
>>> 943 endswitch
>>>
>>> I recall the root object had been called "root_figure" in the recent past and
>>> had been changed to "root" in the 3.1+ branch.
>>>
>>> Can someone verify that get(0, "type") returns "root_figure" and not "root"
>>> for 3.0.4?
>>
>>If you mean get(0).type, I get
>>
>>ans = root figure
>>
>>Marco
>
> ok. The 3.0.x branch needs to have the changeset below added.
>
> http://www.nabble.com/Re%3A--changeset----have-closereq%28%29-respect-the-property-%22tag%22-p20073642.html
>
> http://hg.savannah.gnu.org/hgweb/octave/rev/e41f420875db
>

Unfortunately, it does not apply, and I feel too ignorant about the
graphics code to attempt to resolve the numerous conflicts. A simpler
changeset specifically for 3.0.x would be welcome (if it's worth the
effort).

regards

-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz