[GRASS5] RE: [bug #4140] (grass) Nviz doesn't start either from
GUI or CLI
Hamish
hamish_nospam at yahoo.com
Wed Mar 8 02:55:15 EST 2006
> > Perhaps there's another clue here; when I run nviz from CLI, I get:
> >
> > GRASS 6.1.cvs (Cheticamp):~ > nviz elevation=Cheticamp98.5.ave.fill
> > Loading Data
> > Update elev null mask
> > Loading Data
> > translating colors from fp
> > *Segmentation fault*
> > (my emphasis)
>
> A segfault. I suppose devs would not mind it if you could provide a
> stacktrace. See here for guiding if you don't mind doing this (for
> this purpose, it would be good if your Grass was not built with any
> optimisations nor stripping).
> http://www.qgis.org/index.php?option=com_content&task=view&id=74&Itemid=58
This is not so easy for NVIZ. Or at least it wasn't -- nviz used to be a
script which called a C binary which ran a hybrid TclTk module. Getting
it to run in the debugger was a real chore of setting up the correct
environment. When this bug first cropped up I spent a fair bit of time
trying to track it down without any luck.
My best guess to the cause was a threading issue somewhere in the code
(debian tcl8.4 uses threading while 8.3 doesn't). ???
Bob has since changed things a bit and `which nviz` now calls the binary
directly. Maybe it is time to try again....
configure with:
CFLAGS="-ggdb -Wall" ./configure
.. and recompile using tcl8.4-dev and tk8.4-dev.
I use kdbg as a gdb frontend. Others use ddd, etc
G61 > kdbg `which nviz`
Execution->Arguments
Run with these arguments: elev=elevation
"Program received signal SIGFPE, Arithmetic exception."
Not much use, just tells me it fails somewhere in the Tk_main() call.
Let's try gdb directly:
G61> gdb `which nviz`
GNU gdb 6.3-debian
[...]
This GDB was configured as "i386-linux"...Using host libthread_db library "/lib/libthread_db.so.1".
(gdb) run elev=elevation.dem
Starting program: /usr/local/src/grass/grass61/dist.i686-pc-linux-gnu/bin/nviz elevation.dem
[Thread debugging using libthread_db enabled]
[New Thread 16384 (LWP 7550)]
[New Thread 32769 (LWP 7551)]
[New Thread 16386 (LWP 7552)]
Loading Data
Update elev null mask
Loading Data
translating colors
99%
Program received signal SIGFPE, Arithmetic exception.
[Switching to Thread 16384 (LWP 7550)]
0x42126ae0 in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
(gdb) l
29 *----------------------------------------------------------------------
30 */
31
32 int main(int argc, /* Number of command-line arguments. */
33 char **argv /* Values of command-line arguments. */
34 )
35 {
36 int i;
37
38 Tcl_FindExecutable(argv[0]);
### HB: this is visualization/nviz/src/nvizMain.c
(gdb) bt
#0 0x42126ae0 in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
#1 0x4212667b in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
#2 0x421267bf in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
#3 0x420942b2 in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
#4 0x420942d9 in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
#5 0x42011c5a in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
#6 0x42013d7c in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
#7 0x420141e8 in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
#8 0x4212b1bc in __driUtilCreateScreen () from /usr/X11R6/lib/modules/dri/r200_dri.so
#9 0x41ffd35d in __driUtilUpdateDrawableInfo () from /usr/X11R6/lib/modules/dri/r200_dri.so
#10 0x40853195 in _glthread_SetTSD () from /usr/X11R6/lib/libGL.so.1
#11 0x408532ec in glXCreateContext () from /usr/X11R6/lib/libGL.so.1
#12 0x080628d7 in Togl_MakeWindowExist (togl=0x81b1898) at togl.c:1451
#13 0x0806172f in Togl_Configure (interp=0x807bfb0, togl=0x81b1898, argc=6, argv=0xbfffe778, flags=0)
at togl.c:905
#14 0x08061e7f in Togl_Cmd (clientData=0x808c6e0, interp=0x807bfb0, argc=8, argv=0xbfffe770)
at togl.c:1103
#15 0x406fc66f in TclInvokeStringCommand () from /usr/lib/libtcl8.4.so.0
#16 0x406fd6a4 in TclEvalObjvInternal () from /usr/lib/libtcl8.4.so.0
#17 0x407249c8 in TclCompEvalObj () from /usr/lib/libtcl8.4.so.0
#18 0x40723f0b in TclCompEvalObj () from /usr/lib/libtcl8.4.so.0
#19 0x40757147 in TclObjInterpProc () from /usr/lib/libtcl8.4.so.0
#20 0x406fd6a4 in TclEvalObjvInternal () from /usr/lib/libtcl8.4.so.0
#21 0x406fe254 in Tcl_EvalEx () from /usr/lib/libtcl8.4.so.0
#22 0x40740a24 in Tcl_FSEvalFile () from /usr/lib/libtcl8.4.so.0
#23 0x4070b5ff in Tcl_SourceObjCmd () from /usr/lib/libtcl8.4.so.0
#24 0x406fd6a4 in TclEvalObjvInternal () from /usr/lib/libtcl8.4.so.0
#25 0x406fe254 in Tcl_EvalEx () from /usr/lib/libtcl8.4.so.0
#26 0x406fe6eb in Tcl_Eval () from /usr/lib/libtcl8.4.so.0
#27 0x0805bfac in Ninit (interp=0x807bfb0, w=0x0) at nviz_init.c:459
#28 0x08051884 in NVIZ_AppInit (interp=0x807bfb0) at nvizAppInit.c:130
#29 0x40633f94 in Tk_MainEx () from /usr/lib/libtk8.4.so.0
#30 0x08060f66 in main (argc=3, argv=0xbffff6f4) at nvizMain.c:50
(it failed last year before I was using the radeon card & had dri going...)
still not telling me much. Peppering the C code with printf() and the
Tcl code with puts, I see that it segfaults while in
visualization/nviz/scripts/nviz2.2_script
it helped to turn on debug messages (at the top of nviz2.2_script)
set DEBUG 1
segfaults in
proc Nv_force_panel
on this line:
set Nv_($path) [mk$name\Panel $path]
debug output:
$path=.middle.panelarea.panels.surf
$name=surf
This is called from the Nv_mkPanelMenu proc, "surf" is the first $i
of $Nv_(panels).
changing that line to
set Nv_($path) {mk$name\Panel $path}
gets past that SegFault, but then crashes later on. No idea if {that's}
a correct approach, but it got past the segfault..
Next one: (still in nviz2.2_script)
it gets to
proc Nv_mapPanel
and segfaults on this line:
#insert update to fix KDE wm problem
update
If I comment out "update", it gets past it, but then segaults again a
few lines later on "pack $name -side top -fill x -expand no -anchor n",
or at least a puts after that line never gets displayed.
But on it runs. A window flashes up, I get the "%" prompt on the
terminal, I get past Ninit() and NVIZ_AppInit(), but lose the
scent after nviz2.2_script's init_graphics and Nv_expose procs
finish successfully. (not sure what calls Nv_expose, as it's a bind)
further than I got last time anyway.
Hamish
More information about the grass-dev
mailing list