Tcl NVIZ: load state file "Error: must be surf_id set_mask_mode BOOLEAN"

[hamish]

Hi,

there is a bug in writing the saved-state file in tcl/tk NVIZ. It means that re-loading a state file with raster layers leads to an inconsistent internal state.

The reason is the mask-map inversion status being written as "00" not "0". Fixing the file by hand solves the issue and lets you load the state file correctly.

spearfish example:

GRASS> g.region -d
GRASS> nviz elev=elevation.dem

   File --> Save State
   Save as "test.nviz"

GRASS> head -n 27 test.nviz | tail -n 17
surf*1390794941
map elevation.dem
0
map elevation.dem
0
unset
00
unset
const 60.000000
unset
#888888
0.000000 0.000000 0.000000
3 3
2 2
poly
grid_surf
gouraud


GRASS> nviz state=test.nviz

Loading raster map <elevation.dem@PERMANENT>...
 100%
Loading raster map <elevation.dem@PERMANENT>...
 100%
Translating colors from raster map <elevation.dem@PERMANENT>...
 100%
Diagnostic: Error: must be surf_id set_mask_mode BOOLEAN --
  Load procedure for panel >>>start surf may not be defined

And the internal state becomes a bit messed up: The surface renders but the Raster Surfaces controls show none loaded.

In map_obj.c set_mask_mode() fails because it is saved in the file as "00" not "0" (7th line of the 'surf' entry).

The code that creates it is get_mask_mode() in visualization/nviz/src/map_obs.c, but that is using sprintf(tmp, "%d", mode); to create the string so I'm not seeing where/how the "00" is being introduced, since %d should lead to just "0". ?

The work-around is to edit the .nviz state file by hand and change the "00" to "0". Then the "Load State" works as expected.

Interestingly the first time I launch tcl nviz in 6.4.3 on a different machine it wrote out "0" but later attempts following the exact same series of clicks writes out "00".. smashed stack?

see also scripts/panel_surf.tcl for the reading and writing of the session-state file (Nviz_surf_save() is well commented).

Another question is if after fixing the source of the "00" trouble if we allow sloppy-read, i.e. to auto-translate 00 to 0 upon load so the user can load old+broken .nviz state files created by this bug. Or as a general policy if keeping work-arounds in the code base for old bugs generates too much code-clutter? In this case I would suggest to err on the side of being kind to the victims.

thanks, Hamish

[hamish]

Replying to hamish:

Interestingly the first time I launch tcl nviz in 6.4.3 on a different machine it wrote out "0" but later attempts following the exact same series of clicks writes out "00".. smashed stack?

sorry, my mistake. That happened when switching between 6.4.3 and 6.5, so the bug is in 6.5 only.

Hamish

[wenzeslaus]

6.5 only bug