2.8 KiB
Minutes from 142nd OpenVDB TSC meeting, August 16th, 2022, (EDT)
Attendees: Jeff L., Andre P, Nick A., Dan B., Rich J.
Additional Attendees: Yoshua Nava, Peter Cheng (DW), Karl Marett, Sergio Rojas (Different Dimension), Sebastian Gaida
Regrets: Ken M.
Agenda:
- Confirm quorum
- Secretary
- Siggraph
- VDB 10
- Houdini and VDB 10, PR 1445
- PR 1439
- PR1440
- MSVC Bug
- Next meeting
- Confirm quorum
Quorum is present.
- Secretary
Secretary is Jeff Lait
- Siggraph
NeuralVDB is blasted everywhere. Definitely the most talked about VDB item. Jeff provided his possibly wrong summaries, but we look forward to more details from Ken.
- VDB 10
Matehmatica bindings were promised for VDB 10. But CI and build are needed for it. If we fail at CI, we need at least a third party to install it successfully. Andre can test, but We are missing someone to test on Windows. New projects should always provide CMake to come in. Note Nick will not be around for pre-VDB10-release cmake help.
We don't have any ABI changes. Are there any we want to add? What about Root Node Mask? This pre-dated VDB being open so could be cleaned out. Any virtuals to remove?
We could change leaf iterators to store a pointer to buffer rather than the node. Likewise with the accessors. Avoids the delay-load test.
- Houdini and VDB 10, PR 1445
New warnings show up in NanoVDB with new compilers in NanoVDB, Andre to clean up.
Blosc is bumping to 1.17, so need to add to the containers.
Adding houdini includes causes very weird linker failures. If we can fix this, we can go to C++17.
Both 19.0 and 19.5 fail with these undefined symbols. It could be boost is linking in the houdini lzma. Maybe the core shouldn't be adding houdini dsolibs to the build. Try to bring in tbb/blosc directly without bringing the entire folder? If boost was rpathed directly to its lzma, it might also be fine.
- PR 1439
Fix clamping in vdb visualize sop. Dan is looking for sign off.
- PR1440
Fixes for merge tool in sum mode, and ready for Maximum mode. Again, looking for sign off.
- MSVC Bug
Issue 1446, 1385. This appears to be an MSVC bug. We have a PR to remove boost::any or move to std::any. Perhaps any change will work? Edward's workaround might work.
Can we get a more conformant set of switches? Doing this triggers other incompatibilities though. Setting /permissive- will get us closer to other compilers. Linux and Clang already are default conformant. MSVC defaults to permissive.
This particular problem is a bug with MSVC.
Edward can try a /permissive- on separate branch and see if thre is a simple patch. We could also break it up into component flags and slowly turn on.
- Next meeting
Next meeting is on August 23rd, 2022. 1pm-2pm EDT (GMT-4).