Re: EXC_BAD_ACCESS [was: persistent listeners]

[Ben Abbott]

On Mar 13, 2011, at 1:38 PM, logari81 wrote:

> On Sun, 2011-03-13 at 11:52 -0400, Ben Abbott wrote:
>> On Mar 13, 2011, at 11:31 AM, logari81 wrote:
>> 
>>> On Sun, 2011-03-13 at 11:20 -0400, Ben Abbott wrote:
>>>> On Mar 12, 2011, at 5:06 PM, logari81 wrote:
>>>> 
>>>>> Yes your last change would cancel my intention in my previous change.
>>>>> The idea is that the lines:
>>>>> 
>>>>> ||    if (uimenu->items_to_show ())
>>>>> ||      show_menubar ();
>>>>> ||    else
>>>>> ||      hide_menubar ();
>>>>> 
>>>>> should be before the first allocation of the canvas:
>>>>> 
>>>>> ||      // This allows us to have a valid OpenGL context right away.
>>>>> ||      canvas->mode (FL_DEPTH | FL_DOUBLE );
>>>>> ||      if (fp.is_visible ())
>>>>> ||        {
>>>>> ||          show ();
>>>>> ||          if (fp.get_currentaxes ().ok())
>>>>> ||            show_canvas ();
>>>>> ||          else
>>>>> ||            hide_canvas ();
>>>>> ||        }
>>>>> 
>>>>> otherwise, at least with my fltk version, the first time anything is
>>>>> plotted is vertically shifted by the menu bar height which was very
>>>>> annoying.
>>>>> 
>>>>> So we have to find some other solution. Maybe Shai could give us a hint.
>>>>> 
>>>>> Kostas
>>>> 
>>>> I'm not certain, but it sounds like a problem I have seen on MacOS.
>>>> 
>>>> In my case, I was unable to see the menu bar and the status bar at the 
>>>> same time. In each of these cases, the plot was shifted vertically.
>>>> 
>>>> I was able to fix this by uninstalling fltk-aqua-1.1.10. The package for 
>>>> fltk-aqua-shlibs-1.1.10 remains installed.
>>>> 
>>>> Ben
>>> 
>>> hmm, should I revert the first half of my problematic changeset in hg
>>> until we find a better solution?
>>> 
>>> Unfortunately I do not have a deep understanding of how fltk+OpenGL
>>> works so that I can propose a reliable way for fixing these canvas
>>> initialization issues.
>>> 
>>> Kostas
>> 
>> I checked on Ubuntu, the problematic changset produces a core-dump for ...
>> 
>>      graphics_toolkit fltk
>>      figure (1, "visible", "off")
>> 
>> ... as Soren had expected.
>> 
>> I think it best to revert 12483.
>> 
>> I'm still curious about the vertical shift. I used the simple script below 
>> to verify that I no longer have a vertical shift.
>> 
>>      x = repmat (0:0.1:1, 11, 1);
>>      figure (1)
>>      plot ([x, x'], [x', x])
>>      axis off
>>      set (gca, "position", [0 0 1 1])
>> 
>> With 12483 reverted I don't see a shift.  I've attached a pdf that looks the 
>> same as what is displayed.
>> 
>> Ben
> 
> Now that I have reverted the first half of 12484(9c831d357e6f) your
> crashes should disappear but I get the vertical shift again that you can
> see in the attached pictures. Just after plotting the axes are lower
> than they should (before-drawnow.png). Calling drawnow will bring axes
> to their correct position (after-drawnow.png)
> 
> Kostas
> <before-drawnow.png><after-drawnow.png>

My initial plot looks correct for both MacOS and Ubuntu (I've attached the 
MacOS version). What you're seeing looks similar to what I had experienced when 
I had both the fltk-aqua-1.1.10 and fltk-aqua-1.1.10 packages installed. When 
fltk-aqua was installed fltk-configure did not provide the same ldflag info.

In the event, you may find it useful, on MacOS the correct result is ...

$ fltk-config --ldflags
-L/sw/lib -framework OpenGL -dylib_file 
/System/Library/Frameworks/OpenGL.framework/Versions/A/Libraries/libGL.dylib:/System/Library/Frameworks/OpenGL.framework/Versions/A/Libraries/libGL.dylib
 -L/sw/lib/fontconfig2/lib -L/sw/lib/xft2/lib -L/usr/X11R6/lib -lfltk -lXft 
-lpthread -lXext -lX11

For Ubuntu I see ...

$ fltk-config --ldflags
-Wl,-Bsymbolic-functions -lfltk

Ben