[GRASS-dev] release planning and no volume display on Mac - Any info on where crash happens?
Helena Mitasova
hmitaso at ncsu.edu
Fri Jan 11 12:18:31 PST 2013
On Jan 11, 2013, at 2:48 PM, Anna Kratochvílová wrote:
> Hi Michael,
>
> On Mon, Jan 7, 2013 at 7:57 AM, Michael Barton <Michael.Barton at asu.edu> wrote:
>> So I found a change that would make volumes visible again a week ago. Anyone have any thoughts about this? Do we need the function/procedure that sets the memory pointer for volume displays? Or can we simply get rid of it? Is it needed for Linux? Windows? Any thoughts about what happens without it?
>
> I can confirm that removing those two lines does no harm for me
> (Linux). However it's clear that we cannot just leave them out without
> understanding why they are there. Can anyone (for example Glynn) think
> of a reason why this code is crashing on Mac and not on Linux (no idea
> about Windows)?
I just tested it on Windows 7 with WinGRASS6.4.3RC2 and I am happy to report that it works OK
- both isosurfaces and crossections.
However, in the same WinGRASS6.4.3.RC2 there is a problem with the path to cellheader in G3d_read_window,
so r3.info and g.region rast3d=myvoxel gives an error that the region is invalid.
Also, I finally got a laptop with MacOSX10.8 and I can confirm that in GRASS6.4.3RC2 binary
posted Michael isusrfaces crash wxGUI. I will try to compile myself but I expect the result
to be the same given that it crashed for William as well.
Helena
> And why removing of realloc helps?
>
> Regards,
> Anna
>
>>
>> Cheers
>> Michael
>> ____________________
>> C. Michael Barton
>> Director, Center for Social Dynamics & Complexity
>> Professor of Anthropology, School of Human Evolution & Social Change
>> Arizona State University
>>
>> voice: 480-965-6262 (SHESC), 480-727-9746 (CSDC)
>> fax: 480-965-7671 (SHESC), 480-727-0709 (CSDC)
>> www: http://www.public.asu.edu/~cmbarton, http://csdc.asu.edu
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> On Jan 2, 2013, at 11:22 PM, Michael Barton <michael.barton at asu.edu> wrote:
>>
>>> GREAT NEWS!
>>>
>>> Commenting out both gvl_align_data lines (#684 and #1009) in gvl_calc_c makes this work. Both slices and isosurfaces display and do not crash. I don't know if there would be a problem with larger files or not. But this is finally functional again.
>>>
>>> BUT the command you sent does NOT work. It gives a segmentation fault error:
>>>
>>> m.nviz.image elevation_map=JR_2008_ALL_dem at test3d -a mode=fine resolution_fine=6 color_map=JR_2008_ALL_dem at test3d volume=jr_7408MR_2m_t70 at test3d volume_shading=gouraud volume_resolution=1 volume_position=0,0,20 isosurf_level=1:18.0 isosurf_color_map=jr_7408MR_2m_t70 at test3d isosurf_transparency_value=0 slice=1:x,1:z slice_position=0.000000,1.000000,0.520000,1.000000,0.000000,1.000000,0.250000,0.720000,0.080000,1.000000,0.000000,1.000000 slice_transparency=0,0 position=0.84,0.16 height=356 perspective=20 twist=0 zexag=5.000000 focus=569,593,28 light_position=0.68,0.68,0.80 light_brightness=80 light_ambient=20 light_color=255:255:255 output=jrvoltest format=tif size=798,545
>>> Loading raster map <JR_2008_ALL_dem at test3d>...
>>> 100%
>>> Loading raster map <JR_2008_ALL_dem at test3d>...
>>> 100%
>>> Translating colors from raster map <JR_2008_ALL_dem at test3d>...
>>> 100%
>>> Loading 3d raster map <jr_7408MR_2m_t70 at test3d>...
>>> Segmentation fault: 11
>>>
>>>
>>> Also, somewhere there is a bug somewhere that is sending progress bar text to the screen for any nviz activity (GRASS_INFO_PERCENT: 0...)
>>>
>>> So now we have some kind of workaround. I don't know if there are any other down sides to bypassing this memory reallocation, but this is a good start on fixing this finally. Thanks Anna and Helena.
>>>
>>> Michael
>>> ____________________
>>> C. Michael Barton
>>> Director, Center for Social Dynamics & Complexity
>>> Professor of Anthropology, School of Human Evolution & Social Change
>>> Arizona State University
>>>
>>> voice: 480-965-6262 (SHESC), 480-727-9746 (CSDC)
>>> fax: 480-965-7671 (SHESC), 480-727-0709 (CSDC)
>>> www: http://www.public.asu.edu/~cmbarton, http://csdc.asu.edu
>>>
>>>
>>>
>>> On Jan 2, 2013, at 10:45 PM, Helena Mitasova <hmitaso at ncsu.edu>
>>> wrote:
>>>
>>>> Michael,
>>>>
>>>> what you describe sounds as if you had only one depth layer so let us check first that the 3D region is right.
>>>> If the 3d region is set to the provided 3d raster given here
>>>> http://skagit.meas.ncsu.edu/~helena/grasswork/JR_7408MR_2m_t70.asci
>>>> http://skagit.meas.ncsu.edu/~helena/grasswork/JR_2008_ALL.asci (reference 2d raster)
>>>> it should look like this
>>>>
>>>> GRASS 6.4.3svn (nc_spm_05):~ > g.region -p3
>>>> projection: 99 (Lambert Conformal Conic)
>>>> zone: 0
>>>> datum: nad83
>>>> ellipsoid: a=6378137 es=0.006694380022900787
>>>> north: 250690
>>>> south: 249502
>>>> west: 912791
>>>> east: 913931
>>>> top: 64.00000000
>>>> bottom: 0.00000000
>>>> nsres: 2
>>>> nsres3: 2
>>>> ewres: 2
>>>> ewres3: 2
>>>> tbres: 8
>>>> rows: 594
>>>> rows3: 594
>>>> cols: 570
>>>> cols3: 570
>>>> depths: 8
>>>>
>>>> can you then try to run the following command (please adjust the names of the 2d and 3d rasters to your actual names)
>>>>
>>>> m.nviz.image elevation_map=JR_2008_ALL at Baseline_JR -a mode=fine resolution_fine=6 color_map=JR_2008_ALL at Baseline_JR \
>>>> volume=JR_7408MR_2m_t70 at Baseline_JR volume_shading=gouraud volume_resolution=1 volume_position=0,0,20 isosurf_level=1:18.0 isosurf_color_map=JR_7408MR_2m_t70 at Baseline_JR isosurf_transparency_value=0 slice=1:x,1:z slice_position=0.000000,1.000000,0.520000,1.000000,0.000000,1.000000,0.250000,0.720000,0.080000,1.000000,0.000000,1.000000 slice_transparency=0,0 position=0.84,0.16 height=356 perspective=20 twist=0 zexag=5.000000 focus=569,593,28 \
>>>> light_position=0.68,0.68,0.80 light_brightness=80 light_ambient=20 light_color=255:255:255 \
>>>> output=jrvoltest format=tif size=798,545
>>>>
>>>> it should generate an image with isosurfaces and a tilted crossection - it is not quite what I have on the screen, because
>>>> it apparently uses some default values instead of actual values (e.g. for toggle normal direction and crossection) but it should have some
>>>> isosurfaces. Let us know whether the command line works and we can go from there.
>>>>
>>>> Helena
>>>>
>>>> On Jan 3, 2013, at 12:08 AM, Michael Barton wrote:
>>>>
>>>>> When I put in values between 15 and 20 I can see the horizontal outline of a rectangle but not the 3D shape of a box. No isosurfaces display. No crash either.
>>>>>
>>>>> I un-commented line 684 and commented line 1009 in gvl_calc_c to see what happens to slices
>>>>>
>>>>> /* gvl_align_data(pos, slice->data); */
>>>>>
>>>>> A horizontal slice displays and I can manipulate it in the X and Y direction. But I cannot manipulate it in the Z direction. Also, a Z dimension slice shows up only as a line. Also, isosurfaces do not display. But there is no crash for either isosurfaces or slices when I comment out this line. On the face of it, a memory issue related to display in the Z dimension is causing a crash. Commenting out calls to gvl_align_data in gvl_calc_c stops the crash and stops display in the Z dimension. There is also something called gvl_calc2_c in this folder. Its code is very similar to gvl_calc_c
>>>>>
>>>>> I'm guessing that Martin and Helena know the most about the relevant C code for lib/ogsf/*. Glynn may have some insight into this too. Anyone else is welcome to chime in.
>>>>>
>>>>> FWIW, the last time gvl_calc_c was touched was by Glynn 4 years ago (glynn: Fix char/char* mismatch (merge from trunk, r32757)).
>>>>>
>>>>> Volume display worked as recently as fall 2011 AFAIK.
>>>>>
>>>>> Michael
>>>>> ____________________
>>>>> C. Michael Barton
>>>>> Director, Center for Social Dynamics & Complexity
>>>>> Professor of Anthropology, School of Human Evolution & Social Change
>>>>> Arizona State University
>>>>>
>>>>> voice: 480-965-6262 (SHESC), 480-727-9746 (CSDC)
>>>>> fax: 480-965-7671 (SHESC), 480-727-0709 (CSDC)
>>>>> www: http://www.public.asu.edu/~cmbarton, http://csdc.asu.edu
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On Jan 2, 2013, at 10:21 AM, Helena Mitasova <hmitaso at ncsu.edu> wrote:
>>>>>
>>>>>>
>>>>>>
>>>>>> On Jan 2, 2013, at 1:34 AM, Michael Barton wrote:
>>>>>>
>>>>>>> Testing with GRASS 6.4.3RC2
>>>>>>>
>>>>>>> Rem-ing out line 684 in gvl_calc_c
>>>>>>>
>>>>>>> /* gvl_align_data(dbuff[i].ndx_new, dbuff[i].new); */
>>>>>>>
>>>>>>> prevents crashing. But I don't see an isosurface.
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Helena,
>>>>>>>
>>>>>>> I'm using your test data (jr_7408MR_2m_t70)
>>>>>>>
>>>>>>> What is a good value to set for an isosurface to see anything?
>>>>>>
>>>>>> anything between 15 to 20 should work.
>>>>>> You can move the volume a little bit above the surface to see the entire isosurface
>>>>>> Here is an example of settings that I have used (the name of the volume raster is slightly different
>>>>>> but it is the same data).
>>>>>>
>>>>>>> m.nviz.image elevation_map=JR_2008_ALL -a mode=fine resolution_fine=1 color=243:243:243 volume=JR_7408MR_2m volume_shading=gouraud volume_resolution=1 volume_position=0,0,20 isosurf_level=1:17.0 isosurf_color_map=JR_7408MR_2m isosurf_transp_value=0 position=0.11,0.07 height=243 perspective=13 twist=0 zexag=6.000000 focus=569,593,17 light_position=0.26,-0.30,0.61 light_brightness=94 light_ambient=55 light_color=255:255:255 output=nviz_output format=ppm size=798,545
>>>>>>
>>>>>> This volume includes no-data because the original rasters were masked, I think it would be useful to start with a volume without no-data -
>>>>>> just use r3.null to replace no-data with 0.
>>>>>>
>>>>>> Let me know if it would be useful for me to run same thing in MacOSX 10.6 (I still don't have 10.8 machine quite ready),
>>>>>> (see the slides 12 and 13 here for the isosurfaces at different levels, I have the isosurfaces colored by year - I don't think I gave you
>>>>>> that raster, so your isosurfaces should be just a single color).
>>>>>> http://skagit.meas.ncsu.edu/~helena/publwork/AGU2012lidar6.pptx
>>>>>>
>>>>>> Helena
>>>>>>
>>>>>>> Slices still crash because they still call gvl_align_data. As before, though initially displaying a slice works fine. It only crashes when I try to change something about the slice and it wants to redraw.
>>>>>>>
>>>>>>> Michael
>>>>>>> ____________________
>>>>>>> C. Michael Barton
>>>>>>> Director, Center for Social Dynamics & Complexity
>>>>>>> Professor of Anthropology, School of Human Evolution & Social Change
>>>>>>> Arizona State University
>>>>>>>
>>>>>>> voice: 480-965-6262 (SHESC), 480-727-9746 (CSDC)
>>>>>>> fax: 480-965-7671 (SHESC), 480-727-0709 (CSDC)
>>>>>>> www: http://www.public.asu.edu/~cmbarton, http://csdc.asu.edu
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> On Jan 1, 2013, at 1:24 PM, Anna Kratochvílová <kratochanna at gmail.com>
>>>>>>> wrote:
>>>>>>>
>>>>>>>> Hi Michael,
>>>>>>>>
>>>>>>>> On Tue, Jan 1, 2013 at 8:48 PM, Michael Barton <Michael.Barton at asu.edu> wrote:
>>>>>>>>> Hi Anna,
>>>>>>>>>
>>>>>>>>> On the slim chance that you are online today, can you tell me where gvl_align_data in gvl_calc.c reside in either the source code or compiled code? I have a bit of down time and thought I'd see what I could find out about the volume display breaking. I don't know C but can selectively rem out some things and see what happens.
>>>>>>>>
>>>>>>>> It should be here
>>>>>>>> http://trac.osgeo.org/grass/browser/grass/trunk/lib/ogsf/gvl_calc.c#L685
>>>>>>>>
>>>>>>>> Also, you could try debugging with QtCreator [1] (as an user-friendly
>>>>>>>> interface to the debugger), it is quite easy, here [2] is some help
>>>>>>>> for setting up the project. I don't know what method you would like to
>>>>>>>> use but this is the easiest I know. I suppose it should work on Mac,
>>>>>>>> too.
>>>>>>>>
>>>>>>>> Regards,
>>>>>>>> Anna
>>>>>>>>
>>>>>>>> [1] http://qt-project.org/downloads
>>>>>>>> [2] http://grasswiki.osgeo.org/wiki/Using_QtCreator_for_GRASS_C_development#Debugging
>>>>>>>>
>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Thanks
>>>>>>>>> Michael
>>>>>>>>> ____________________
>>>>>>>>> C. Michael Barton
>>>>>>>>> Director, Center for Social Dynamics & Complexity
>>>>>>>>> Professor of Anthropology, School of Human Evolution & Social Change
>>>>>>>>> Arizona State University
>>>>>>>>>
>>>>>>>>> voice: 480-965-6262 (SHESC), 480-727-9746 (CSDC)
>>>>>>>>> fax: 480-965-7671 (SHESC), 480-727-0709 (CSDC)
>>>>>>>>> www: http://www.public.asu.edu/~cmbarton, http://csdc.asu.edu
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On Dec 6, 2012, at 12:07 AM, Anna Kratochvílová <kratochanna at gmail.com> wrote:
>>>>>>>>>
>>>>>>>>>> Hi Michael,
>>>>>>>>>>
>>>>>>>>>> On Wed, Dec 5, 2012 at 11:02 PM, Michael Barton <michael.barton at asu.edu> wrote:
>>>>>>>>>>> Any suggestions yet on where the crash I documented in
>>>>>>>>>>> http://trac.osgeo.org/grass/ticket/1736 actually happens in the code so that
>>>>>>>>>>> I can take a look at it and see if there is something fixable for the Mac? I
>>>>>>>>>>> posted what I hope is the relevant debug output to help identify this.
>>>>>>>>>>>
>>>>>>>>>>> While we may need to think about wxPython 2.9, it seems best to first look
>>>>>>>>>>> at what line is actually causing the crash and see if there is a fix or
>>>>>>>>>>> workaround.
>>>>>>>>>>
>>>>>>>>>> The crash occurs probably in gvl_align_data in gvl_calc.c. The error
>>>>>>>>>> message 'pointer being realloc'd was not allocated' is quite clear,
>>>>>>>>>> however it's not clear why it happens on Mac only. Maybe someone with
>>>>>>>>>> better knowledge of C could understand it more.
>>>>>>>>>>
>>>>>>>>>> Anna
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Michael
>>>>>>>>>>> ____________________
>>>>>>>>>>> C. Michael Barton
>>>>>>>>>>> Director, Center for Social Dynamics & Complexity
>>>>>>>>>>> Professor of Anthropology, School of Human Evolution & Social Change
>>>>>>>>>>> Arizona State University
>>>>>>>>>>>
>>>>>>>>>>> voice: 480-965-6262 (SHESC), 480-727-9746 (CSDC)
>>>>>>>>>>> fax: 480-965-7671 (SHESC), 480-727-0709 (CSDC)
>>>>>>>>>>> www: http://www.public.asu.edu/~cmbarton, http://csdc.asu.edu
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>
>>>>>>>
>>>>>>
>>>>>
>>>>
>>>
>>
More information about the grass-dev
mailing list