# 8/27/2020

### Attending:

* Cary Phillips
* Christina Tempelaar-Lietz
* Joseph Goldstone
* Kimball Thurston
* Nick Porcino
* Owen Thompson
* Peter Hillman

### Discussion:

* Imath release:

  * Need a beta period. Larry: I tried to use master and fell on my
    face.

  * Legacy headers: Imath headers have traditionally been installed in
    OpenEXR subfolder, and existing user code has:

        #include <OpenEXR/ImathVec.h>

  * Downstream packages need to work cleanly with both current
    installation of v2.* and with 3.0.

  * Proposed solution: adaptive headers. For some number of upcoming
    releases, we will provide stub headers in OpenEXR folder, and
    issue warnings. Then in a future release, remove them. This will
    aid in user transition.
  
  * Should be no more FindOpenEXR.cmake.

  * Should Python on by default? The majority the complexity is in
    handling both 2 and 3.  Kimball: could specify a path to the
    python binary.  Larry: propose off by default, so it builds
    better. Everyone agrees.

  * Is Half in a namespace? Only in CUDA. It in the namespace, with an
    optional using. CUDA forces the option to true.

* Versioning policy: we will bump version number immediately *after* a
  release, so that building from the dev branch produces library names
  with the not-yet-released version, not with the current version that
  it doesn't, in fact, match.

* Fuzz: damaged OpenEXR files can cause Undefined-shift operators in
  B44 uncompress (#821).

* Apple hardware: 

  * Wave was talking at a TAC meeting.

  * Apple has some optimizations, but waiting until silicon is available.

  * Larry: we should make the half stuff work with Intel hardware,
    faster than table-based conversions.

  * DTK: Some people are getting a working box. Nick has one on the
    way, will do a dry run to make sure USD works. Take OpenEXR for a
    spin, too. It’s a mac mini crammed full of hardware.  Probably
    rigged it up to run an ipad arm chipset.