bug in Anatomist 4.3.0 64bit (on Ubuntu 12.04LTS 64bit)

[rvboutin]

Hi,
I have got a bug consistently happening when I try to draw manually ROI on a CT image (size 480x480x639, resolution 0.204x0.204x0.204mm). After opening the image and dragging it to a view, I create a new ROI session and a new ROI, after a number of drawing (variable from one session to another) Anatomist hangs with the following error:
...

Code: Select all

notify change for Brain (2)
notify change for Brain (2)
*** glibc detected *** /home/herve/brainvisa-4.3.0/bin/real-bin/anatomist: realloc(): invalid next size: 0x00000000059a2c80 ***

Obviously, the address refereed to by "realloc(): invalid next size:" changes everytime. At this point there is nothing else to do that kill the task.


For info, on whatever version of Anatomist I use, I got the following log in the terminal when starting:

Code: Select all

herve@Home-PC-Ubuntu:~/brainvisa-4.3.0/bin$ ./anatomist 
Fontconfig warning: "/etc/fonts/conf.d/11-lcd-filter-lcddefault.conf", line 9: invalid constant used : lcddefault
Fontconfig warning: "/etc/fonts/conf.d/53-monospace-lcd-filter.conf", line 17: invalid constant used : lcdlegacy
Starting Anatomist.....
config file : /home/herve/.anatomist/config/settings.cfg
PyAnatomist Module present
global modules: /home/herve/brainvisa-4.3.0/share/anatomist-4.3/python_plugins
home   modules: /home/herve/.anatomist/python_plugins
loading module modelGraphs
loading module selection
<H1>Anatomist launched</H1>
loading module gradientpalette
loading module meshsplit
loading module paletteViewer
loading module volumepalettes
loading module bsa_proba
loading module anacontrolmenu
loading module foldsplit
loading module histogram
loading module palettecontrols
loading module profilewindow
all python modules loaded
Multitexturing present
Number of texture units: 8
GL_ARB_shadow present
GL_SGIX_shadow extension not available
GL_SGIX_depth_texture extension not available
GL_ARB_depth_texture extension present
GL_ARB_texture_cube_map extension present
GL_EXT_texture_cube_map extension present
Number of texture units: 8
Unrecognised OpenGL version

I hope this help.
NB: the bug does not happen with the 32bit version of Anatomist, and when starting the 32bit version the log stops at: "all python modules loaded", the following lines about GL_etc... are not there.
Cheers,
Rv

[riviere]

I am trying to draw on a large volume (with different resolution however) using anatomist 4.3.0 64 bit, and after a while I don't reproduce the problem. But as I am only drawing until il crashes, maybe there are some operations that you do when actually working with ROI drawing and which I am not doing right now. Can tou remember if there are specific conditions when it occurs: while drawing, or erasing, or filling a closed region ? Do you have several ROIs ?
Or maybe I have not used it long enough to trigger the problem... (I gave up after a few minutes)
Or could it be related to memory usage ? The 64 bit version consumes more memory than the 32 bit one, and you might have run out of memory ? The ROI tool remembers all dawing events to allow undo, and as far as I remember, we never took care of limiting this undo buffer, so with such a large image, it may grow quite large (which is something we should fix, anyway).
Denis

[riviere]

By the way, the GL... messages only appear when you open your first window in Anatomist.
Denis

[rvboutin]

Hi Denis,
OK here is the procedure:
1. I load the image in Anatomist, and open a view;
2. Create a ROI session, create 1 ROI and start drawing;
3. I keep drawing until Anatomist hangs. What I draw (shape, etc..), where on the image, etc.. does not affect the bug; the bug happen after some times/number of drawings

I think you might be right about the number of operations stored in memory, although I was always far from reaching the maximum amount of RAM (I have between 12 and 16Gb of RAM on all PC) and on the recent trials I did to reproduce the bug I was always using about 25% of the RAM at max.

Here is the full log of one session (but from memory, apart the number of drawing that varies each time, there are always identical):

Code: Select all

herve@Home-PC-Ubuntu:~$ cd brainvisa-4.3.0/bin/
herve@Home-PC-Ubuntu:~/brainvisa-4.3.0/bin$ ./anatomist 
Fontconfig warning: "/etc/fonts/conf.d/11-lcd-filter-lcddefault.conf", line 9: invalid constant used : lcddefault
Fontconfig warning: "/etc/fonts/conf.d/53-monospace-lcd-filter.conf", line 17: invalid constant used : lcdlegacy
Starting Anatomist.....
config file : /home/herve/.anatomist/config/settings.cfg
PyAnatomist Module present
global modules: /home/herve/brainvisa-4.3.0/share/anatomist-4.3/python_plugins
home   modules: /home/herve/.anatomist/python_plugins
loading module modelGraphs
loading module selection
<H1>Anatomist launched</H1>
loading module gradientpalette
loading module meshsplit
loading module paletteViewer
loading module volumepalettes
loading module bsa_proba
loading module anacontrolmenu
loading module foldsplit
loading module histogram
loading module palettecontrols
loading module profilewindow
all python modules loaded
Multitexturing present
Number of texture units: 8
GL_ARB_shadow present
GL_SGIX_shadow extension not available
GL_SGIX_depth_texture extension not available
GL_ARB_depth_texture extension present
GL_ARB_texture_cube_map extension present
GL_EXT_texture_cube_map extension present
Number of texture units: 8
Unrecognised OpenGL version
ToolBox::updateActiveControl, this=0x445e400, descr: Default 3D control
Free Framework
newGraph: roiName: a01168_01_20121211_v1.ct_ROI.arg
bounding box found : 0, 0, 0
                     479, 479, 638
Label Volume Dimension : 480, 480, 639
ToolBox::updateActiveControl, this=0x445e400, descr: PaintControl
notify change for brain (2)
OpenGL error tex: invalid framebuffer operation
TexImage failed
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
~GLTexture - Bad GL texture - 0
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
notify change for brain (2)
*** glibc detected *** /home/herve/brainvisa-4.3.0/bin/real-bin/anatomist: realloc(): invalid next size: 0x000000000c738140 ***

When it crashed, the memory usage was: 3360Mb/11981Mb, swap 0. It crashed after 42 modifications.
I forgot to say that it also happens with PET file that are 128x128x159 voxels @ 0.75mm resolution.
EDIT: it happens with dynamic PET files (I tried a 35 frames one this morning, so 350Mb large) and it crashed. I just tried with a sum PET image 5Mb, and it has not crashed after 342 modifications of the ROI... I am giving up...
Hope this helps.
Rv

[rvboutin]

Hi Denis,
I don't know if you have had time to look into this, but the bug is still happening... I can't find any reasons for that.... when I said "I am giving up", I meant that I stop reproducing the bug with the sum PET image after 342 modifications... not that I stopped looking into the problem.
The problem does not happen on specific drawing action (filling up, closing, etc..), it just happens after a random number of modifications to 1 ROI.
Hope you can solve this bug somehow...
Cheers,
Rv

[riviere]

Hi Rv,
No I could not reproduce the problem so far. Maybe I have not been patient enough when clicking and clicking. I'll try again when I get time.
Denis