LoadVBX bug (not a SaveVBX bug)

NVIDIA / gvdb-voxels

Sparse volume compute and rendering on NVIDIA GPUs

Other

679 stars 145 forks source link

LoadVBX bug (not a SaveVBX bug) #19

Closed coreuml closed 4 years ago

coreuml commented 6 years ago

In SaveVBX function, at line 1814 if (major >= 2) { fwrite( &use_bitmask, sizeof(uchar), 1, fp ); // bitmask usage (GVDB 2.0 or higher) }

It 's not writed bitmask sign, but in LoadVBX, read this sign.

Modified: if ((major == 1 && minor >= 1) || major > 1) { fwrite( &use_bitmask, sizeof(uchar), 1, fp ); // bitmask usage (GVDB 2.0 or higher) }

ramakarl commented 5 years ago

Yes, that looks like a bug. The condition in SaveVBX is the correct one (only 2.0 and higher). If you make a pull request with LoadVBX fixed to match it then i will merge. Also, LoadVBX should say 'fread'

ramakarl commented 5 years ago

Just to clarify whats going on here.

In GVDB 1.0 the bitmask was required internally.
In GVDB 1.1 the bitmask is not used. There is no option to use it.
In GVDB 2.0+ it might be possible to use either, and this option would be saved to vbx files only in 2.0 and higher.

The logic is: If 1.0.then use_bitmask=true; If 1.1 then use_bitmask=false; If >=2.0 then read/write use_bitmask to disk;

NBickford-NV commented 4 years ago

I've just pushed commit c530d1aa, which should fix this issue - it makes it so that if LoadVBX reads a version number of 2.0 or greater, then it loads use_bitmask, while if it reads a version number of 1.0, then it sets use_bitmask to true, and otherwise it sets use_bitmask to false. It also updates the VBX file specification to include these changes between different versions.