TSC Meeting Notes 2024-05-02
Attendance:
Discussion:
Python bindings:
It should allow reading of the header only
All parts supposed to have the same header? There are assumptions/restrictions that some of the metadata is the same across all parts.
Timecode is supposed to be the same across parts. Parts aren’t frames in a sequence
Larry: I would love to lift restrictions
PR 1734 changing implementation of unions
Objecting to warningsÂ
Kimball: we require C11
Issue 1737 Homebrew issue
Tell formula to install libdeflate before openexr
Cary: curious which version they are picking up
Imath PR 397: Problem with alembic, Lucas
Reverting a previous fix, would this reintroduce the previous issue?
Larry: original fix should have been to get from site package
Change the cmake module file that is installed, rather than look for library in /usr/lib, import directly from site package
Ask Lucas to test this
PyPi Imath projectÂ
Cary: someone created package there but it doesn't work and contact on the project has not replied, filed request with packaging authority to take ownership, dozens of these requests so not sure if we can get to a point where we can provide 'pip install imath'
Nick: we had this issue on usd so have 'pip install usd-core'Â
Nick: should make imath-aswf , something official like imath-official
Cary: attaching it to openexr somehow? But that goes against logic of splitting it off
Larry: are all legit uses of imath in the context of openexr?
No mechanism to distinguish the official version
Kimball: if we use -offical with imath we should do it with exr tooÂ
Cary will make another pass over it
Slack channel post re: metadata and long list of issues
Cary: hodgepodge of issues, request for 16-bit integers
Kimball: discussion with Peter re lower or higher-bit integers - type is uint but has a bit resolution in it, easy to make it backwards compatible
Nick: have to couple that with an encoding, say what those 32 values are for example, table saying how it is quantized
Kimball's recent changes
Next PR will be exposing functions that exist internally in encode/decode pipeline as first class functions so people can use them, also allows to remove all the compression code from the Imath library. Passes through to the C layer. Will clean up majority of security code quality reports. Will be interested to see if it autodetects that we fixed it.
Nick has been reviewing keeping in mind how it might affect USD.
Kimball: pulling back what was originally going to do, no downstream applications should have to change anything. If they do, that's a bug. The ABI of the classes will change.Â
Nick: all changes look very useful.
Kimball: weird interplay between utility objects that are exported. Similar things in core library but they are not exported. Preserving those but normally would want to delete it since we aren't using.Â
Nick: could start marking them as deprecated. Could target the next release for that.Â