2022-10-17
October 17, 2022
Host: Carol Payne
Secretary: Rémi Achard
Attendees:
Apologies:
OCIO TSC Meeting Notes
OCIO Archive PR
Doug: First 2 CI tests on Linux fails, Docker VFX2019 use an older CMake version not compatible with minizip-ng build.
Doug: Different options here: 1/ try to patch minizip-ng, don't seem good idea, 2/ update the docker container used to the minimum version needed, 3/ phase out VFX2019 and move over to upcoming VFX2023.
Carol: VFX2019, 4 years old, might start to get tricky to maintain, we should think about dropping it.
Kevin: Should we try to update CMake at the start of every CI job? Also mentioning VFX Industry spreadsheet (see below) suggest some still depend on old software versions.
Michael: Updating CMake version used by the CI jobs seem like a good idea.
Remi: VFX2023 going to release soon, we should think about which jobs to drop in the matrix to include it if we want to keep the 20 jobs limit.
Michael: Can update the Docker image relatively easily because it is controlled by the OCIO image from the Docker repository.
Doug: Will sync with Michael on this.
MinGW Local Fix PR
Carol: Can someone review it?
Doug: Not sure it will make 2.2 release but should be fine for 2.2.1
Kevin: Really small patch
Roadmap discussion
Doug: Release by October 31 to be part of VFX 2023. Plan is to have 2.2.1 near the end of the year, opportunity to include remaining fixes, including for bug found in 2.2.0 tests.
Doug: Focusing remaining 2.2.0 work on new feature to lock the API, minor enhancement for 2.2.1. Lot of things going on in parallel at Autodesk, small delays expected.
Carol: Area where folks can jump in and help on the remaining work?
Doug: Will think about it, thanks.
Mandatory interchange roles PR
Doug: Discussion on GH issue, talks at Siggraph this August. Plan is to have aces_interchange and cie_xyz_d65_interchange mandatory in 2.2.
Doug: PR ready, include some update to ociocheck and the config validate() method. Validation will warn but not fail in case of missing roles.
Doug: One reason for that, when you update a v1 config with upgradeToLatestVersion() it is non trivial to add the missing roles when this is done in the context of an end-user application, a warning will be logged but the validation will not fail. Similar behaviour for a couple of other roles including scene_linear, color_timing (used in LegacyViewingPipeline) and compositing_log. These check are only performed for 2.2 + configs.
Doug: Please see PR for more details and provide your feedback. Open to discussion about potential role to add or remove to the check list.
BuiltinTransform
Doug: PR for LogC4 and CLog2 and CLog3 should come up later this week.
Colour space isLinear and conversion from/to external spaces
Doug: PR for 1399 (isLinear) and 1541 (convert from/to external colour space) issues in progress. Hope to have that for next week, not a lot of code, should be quicker to review than archive for example.
Doug: Challenging to have these in time for 2.2, if too late or people want more time for review, can push away but would be for 2.3, not possible to include in 2.2.1 du to ABI / API changes.
Carol: Understandable, would be great to have but not a deal breaker I think.
Hash PR
Doug: Asking about the Hash PR, it's been open for months, think the group should reach a decision. Don't change the API but substantial enough update, should discuss if still wanted for 2.2.0.
Doug: Issue raised with it is that it bring a dependency for the new hash function, any thoughts?
Remi: User facing change now that the config hash is printed in ociocheck.
Carol: Will ping Mark B. and ask for one last round of review / comment.
Doug: Potentially lot more work if we want to use custom implementation or less well known implementation.
Kevin: We are not cryptographic expert, xxHash tends to come up first when you search for non cryptographic hash function and seem to be the go to solution.
Michael: Embedded internal dependency, should not require lot of maintaining.
Remi: Remaining point about symbol leakage check, will have a look this week and share conclusion on the PR.
NamedTransforms PR
Michael: NamedTransform overloads for getProcessor, API change, please review.
HLSL Fix PR
Michael: Please review, Epic in need of the fix.
Misc PRs
Doug: Going to review these and the ICC profile PR this week.
Remi: Number of issues raised related to GradingPrimary, candidate for 2.2.1 bug fix release.
Doug: Small PRs from Remi, quick to review, please have a look.
Branch and Release management
Doug: Will create RB- branch for 2.2 soon. Usually restriction for PR's 2 reviews not coming both from Autodesk but should be lifted for release branches.
Michael: Confirming RB- branch only require 1 approval now.
Carol: We have both a new version of the library and new ACES configs at the same time. Will check with Emily about a blog post on ASWF site.
Carol: Planning for a separate blog post about the general OCIO config usage and workflow, what the new changes mean (eg. BuiltInConfig)
Kevin: Worth discussing with DCCs vendor about our intentions on 2.2.0 and 2.2.1, ideally should use directly 2.2.1, 2.2.0 is only for ABI stable VFX 2023 release and "good" version is 2.2.1.
Carol: Similar to what we did last time, but good to mention. Mark what do you think?
Mark: In theory we should target VFX2023, so we should use the latest patch release. Planning phase to begin, will have more informations soon.
Carol: Plan to send an email to main DCCs vendors to discuss about our release process and differences between 2.2.0 and 2.2.1 releases.
Kevin: Using the ReferenceGamutCompress BuiltIn as an example, mix of DCCs supporting it or not, hard to track and rely on. Ideally we should avoid mixture of versions, especially in case bug are found.
Mark: From patch release point of view, we always push to for the latest patch release, don't think we would use 2.2.0 if 2.2.1 is available.
Kevin: Some vendor are slow to adopt latest patch releases.
Carol: Will compile list of vendor and reach out to discuss our release process, will sync with the group before proceeding. Plan for more general outreach program to developer implementing OCIO in their applications.
Config WG Update
Carol: Planning to release new RC version tomorrow. Changes in alia, categories and naming among other things. Also fix issue with missing BlackMagicDesign colour space found by Doug. Un-Tonemapped view added to the list of active views.
Carol / Doug: Roles to include and values, will be discussed and finalised in the WG meeting tomorrow. Plan is to remove default, reference and rendering roles but otherwise keep the role from ACES 1.2 for consistency reasons. Removing the default role is a good step to encourage use of FileRules.
Remi: USDview using default role for DisplayTransform input space, PR in progress for OCIO 2, could raise the issue over there.
Carol: Please take a look at the new configs when they are released if you have spare cycles, good to have new eyes on it.
Carol: Technically last WG before OCIO 2.2 deadline.
VFX Industry Matrix Link
Carol: You should have ability to view and comment, if anyone want to help filling in the info, please contact me.
Carol: It provides a good overview of what other projects are doing. Goal would be to start aligning on version dependency among other things.
Carol: If you have any questions, please reach out to me or Larry directly.
OFX Plugin discussion
Mei: Checked with Remi and Michael, able to get OFX plugin to show up in Resolve. But as soon as OCIO config env var is set (absolute path), it stop working. Remi and Michael have it working, not sure how to progress from there.
Michael: Could it be a compatibility issue between OCIO library and config version?
Mei: Don't think so, I'm getting latest build from main and tried a number of OCIO config, from ACES configs 1.0.3 to 2.X but none of them work. As soon as the OCIO env var is set, controls disappear and the plugin do not work.
Michael: Should look for Resolve error log.
Doug: Should also check if ocioconvert would work with the config used.
Remi: Should try using debug prints in the OCIO OFX plugin loading code sequence?
Michael: In Resolve, in the Help menu / Create diagnostic option. Might need to hold option / control depending on the platform to see it. Will dump a big text file with the error log, hopefully should see an error.
Mei: Will try these thanks, also mentioning I'm using M2 Apple hardware and the latest Resolve version.
Carol: Should try to document your findings, anything we could share to improve the workflow? OFX is now an ASWF project, to keep in mind.
Slack History
Doug: Free trial ending soon, history might disappear in a few days.
Kevin: Thomas dump archive periodically on colour-science slack.
Carol: Will check with Thomas if I can use that workflow on the OCIO slack instance.
Kevin: https://slack.com/intl/en-gb/help/articles/201658943-Export-your-workspace-data worth looking at, doesn't include private conversations (which is good).
Carol: Will look, end of trial on 20th this week.