TSC Meeting Notes 2021-11-18
Attendance:
Discussion:
Follow-up from Richard Shaw’s TAC presentation (Fedora package maintainer):
If we had it to do over again (3.0 release): understand our downstream consumers
We should find what versions the standard distributions are on and encourage them to upgrade.
Kimball: I did that for v2.4
Larry: a good place to start: https://repology.org/project/openexr/versions
How do you find who the maintainers are?
Kimball: for Ubuntu, go into the package distros, it shows the maintainers
Kimball: Sometimes patches are useful/interesting, hard to discover that there are patches.
Cary: we should also meet with build engineers at Foundry, SideFX, autodesk.
Kimball: should we mutate the reference platform going forward?
Original motivation: encourage everyone to move forward
DCC’s are moving forward at different paces
Problem with different DCC’s doing things in a different way (with/without namespaces)
Don’t know which ones use libraries internally, vs. leading symbols.
Outstanding fuzz issues:
Peter: primary outstanding ones are due to running out of memory. Deep allocates memory up front. Can open a deep image with the scanline interface.
A small file can allocate a huge amount of memory.
The test should take the same steps that pedantically written code should.
Don’t want the test to reject things that are wide.
Once it gets beyond a certain size, it pre-allocates.
Allocating a huge amount of memory up front might not be faster.
Should allocate along the way
RgbaInputFile PR
Should the RgbaInput file should be quite limited. You get the yuv2rgb conversion
Runpath issue #1204:
Line in OpenEXRSetup.cmake: “set runpath to ../lib”
So the executables have the expected lib
Kimball: sould use tmpsyspath
Kimball will take a look
Kimball: we should push the global clang-format reformat.
Cary: Should rename master to main soon. Will send an announcement to openexr-dev.