[katana] unexpected geometry.point.P sample times when using a reference and offset

[nxkb]

Description of Issue

When rendering a reference with an SdfLayerOffset, the attribute data that winds up in geometry.point.P sample times is not what we expect.

The sample times are -.25, 9.75, 10, 10.25, when we expect something more like -.25, 0 -.25. Or possibly 9.75, 10, 10.25.

Steps to Reproduce

1. Open attached katana file and render at any frame > 10

System Information (OS, Hardware)

Linux

Package Versions

0.8.2

referenceOffsetFiles.zip

[spiffmon]

Thanks for filing, Nick - we'll take a look!

On Wed, Dec 13, 2017 at 6:44 PM nxkb ***@***.***> wrote: Description of Issue When rendering a reference with an SdfLayerOffset, the attribute data that winds up in geometry.point.P sample times is not what we expect. The sample times are -.25, 9.75, 10, 10.25, when we expect something more like -.25, 0 -.25. Or possibly 9.75, 10, 10.25. Steps to Reproduce 1. Open attached katana file and render at any frame > 10 System Information (OS, Hardware) Linux Package Versions 0.8.2 referenceOffsetFiles.zip <https://github.com/PixarAnimationStudios/USD/files/1557786/referenceOffsetFiles.zip> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#352>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AF7qaGSORRepj-HTzKCsKYlcfQ9tPCyhks5tAIt3gaJpZM4RBd8O> .

-- --spiffiPhone

[spiffmon]

Hmmm... it looks like we missed a codesite or two that needed updating when we un-inverted the layerOffsets. I can repro bad behavior in your test just using GetTimeSamplesInInterval(). Sorry for the trouble - I'm filing an internal bug specifically for the core issue, but let's leave this open for now just in case there's additionally anything to fix in Katana (hopefully not).

In the meantime, I confirmed that the old behavior still works as expected, which you can get back to, for now, by setting the env var USD_USE_INVERSE_LAYER_OFFSET=1

--spiff

[jtran56]

Filed as internal issue #155178.