support onetbb and boost 1.87.0 #1458
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
They don't match our set semantics, and typically generate huge numbers of warnings during loading. We only ignore them on UsdLuxLights (where they are built in to the base schema), so users can still use those names for collections/sets elsewhere.
…lections USDScene : Don't treat `lightLink` and `shadowLink` as sets
The remapping discussed appears to be performed for us inside `UsdSkelSkinningQuery::ComputeSkinnedPoints()` and `UsdSkelSkinningQuery::ComputeSkinnedNormals()`, as well as in our new facevarying normals code. That would explain why an attempt to apply it manually would lead to things getting jumbled up.
…Skinning USD PrimitiveAlgo : Fix loading of facevarying skinned normals
In this case, `UsdStage::CreateNew()` returns a null result, which we were dereferencing, leading to a crash. We now just throw an exception describing the problem instead.
USDScene : Fix crash trying to write to an open stage
This was broken by 8ead0c7, which inlined the functions, meaning that client code inlined the version numbers at the point of compilation. There's no point in doing that - that's what the macros are for.
Version : Fix `*Version()` functions to return runtime version
On Windows, the result of USD's `GetResolvedPath()` method returns the resolved path using `\` between directories. This can cause problems when doing string substition on a resolved path since the `\` will be interpreted as the start of an escape character. Using `/` also conforms to Cortex's standard of always using `/` for paths because they work on all operating systems.
If an attribute has no value, `DataAlgo::fromUSD()` returns `nullptr`, and we were dereferencing it. We could probably still use a `static_cast` here as long as we checked for null, since we don't expect anything but StringData when non-null. But since I failed to predict this outcome, I've gone belt-and-braces and used a `runTimeCast()`.
USD VolumeAlgo : Fix crash loading empty field
- Add stage timecode range, so we get a playback range in `usdview`. - Rename `Skeleton` prim to "skel", because having it share a name with one of its joints was confusing.
The call to `skinningQuery.GetJointMapper()` can not only reorder transforms but can also omit them. This occurs when SkelBindingAPI uses a `skel:joints` attribute to bind a mesh to a subset of joints, allowing the weights for other joints to be omitted. We were assuming that the mapped transform array had the same length as the input array, and were running off the end of the array.
…ngFix USD PrimitiveAlgo : Fix crash loading skinned facevarying normals
We're not using the result anywhere. I assume this was an early attempt at ignoring unauthored values, eventually replaced by the code below.
IECoreUSD::PointInstancerAlgo : Support inactive / invisible ids
This is needed when values aren't sourced from attributes - for instance, from metadata instead.
I think we should be moving away from pre-registered metadata entirely, since the definitions will not be available anywhere outside of Gaffer.
Nothing in the code is changing here - we're just putting related stuff in the same part of the file.
The main motivation here is that the MaterialX shaders we were using as adaptors don't really make sense - they're not actually available as OSL shaders any more, so they have zero chance of working in other DCCs when we export shading to USD from Gaffer. Instead we can now register renderer-specific adaptors that are available everywhere the renderer is. This also somewhat changes our approach to removing adaptors. Instead of just tagging them as "an adaptor" and figuring the rest out based on the shader name, we now register all the necessary information as blind data. This will allow us to reliably remove an adaptor even if adaptor registrations change in the future. It should also be useful at a certain studio which rewrites the `info:id` attribute on export to USD, in such a way that it broke our old adaptor removal code.
…ments ShaderNetworkAlgo improvements
CI : Update for the Node20 future
MeshAlgo::MeshSplitter : Preserve vertex order
Xcode 15's new linker now complains when asked to link to the same library multiple times. I've removed the duplicates from the modules required to build a regular Cortex release, though I haven't touched the DCC specific modules. If the existing duplicate library specification turns out to be necessary for other platforms, we could alternately disable the linker warning here with `-no_warn_duplicate_libraries`. The new linker also complains that the `-single_module` flag is obsolete. `-single_module` became the default behaviour way back in OSX 10.4, so we should be fairly safe to remove it here. Disabling FMA via `--ffp-contract=off` fixes a test failure with `MeshPrimitiveEvaluator.pointAtUV()` returning an incorrect value in `TestMeshPrimitive.testPlane()`. Curiously, even with `--ffp-contract=off`, this test still fails when run on a virtualized instance of macOS. GitHub's macOS runners are vitualized, so we'll end up skipping this test when we enable macOS CI...
These builds lack signing and notarization and are published chiefly for our own use. Making the build artifacts available, even in their current state, paves the way towards re-establishing Gaffer CI runs on macOS.
Reverted recent update since ImageEngine#1441 made it unnecessary.
This is needed for the tests to locate the DLL binaries that are currently in both `bin` and `lib` subdirectories. When all DLLs are moved to `lib`, this can be reverted.
`repr` Infinity
Adds option to server to allow a 'merge' driver that shares a display driver to write to.
…rToMergeDrivers Update DisplayDriverServer
…tsAgain IECoreHoudini : reverts 9ead39d
Hopefully this makes it a little bit clearer what is being returned.
USDScene animation fixes
Update to Dependencies 9.1.0
If you set the env var IECOREUSD_POINTINSTANCER_RELATIVEPROTOTYPES=1, if an instancer named "/inst" contains an internal prototype path like "/inst/Prototypes/proto", that prototype path will now be loaded as "./Prototypes/proto". This is compatible with how point instancers will now be handled in Gaffer, allowing them to be relocated within the scene hierarchy.
IECoreUSD : Load prototype paths inside a PointInstancer as relative
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Generally describe what this PR will do, and why it is needed.
I added support for onetbb (Tested with onetbb 2022.0.0)
I added support for boost 1.87.0 (Tested with boost 1.87.0)
Related Issues
Dependencies
Breaking Changes
I added support for onetbb (Tested with onetbb 2022.0.0)
I added support for boost 1.87.0 (Tested with boost 1.87.0)
I tried to test with the test suite, but I don't know which python library to install (imath python library)
Checklist