5.6.0 - feature freeze

[TurboGit]

The current plan is to feature freeze on May 15th. Please let me know what are the features cooking that you'd like in 5.6.0.

[jenshannoschwalm]

Proposal to accept neg rgb data while interpolating

[TurboGit]

@wpferguson @andriiryzhkov @zisoft @kofa73 @dtorop @ralfbrown @masterpiga @kmilos @LebedevRI

Just pinging you to ensure you see this announce. Nothing more.

[zisoft]

#18468 is marked for 5.6 and ready from my side

[TurboGit]

@zisoft : Good I asked because I thought I had already merged this. Will do that right away.

[andriiryzhkov]

On my side, two big pieces left:

1. One important piece is left on GPU acceleration installation in PR [AI] Add ONNX Runtime GPU acceleration install flow to AI preferences #20963.
2. Also Lua AI API in PR [AI] Add Lua AI API with model inference, tensor ops, and raw CFA support #20716. @wpferguson I am waiting on your review and opinion on if this is ready to be included in release.

Also two issues I am working right now (not new features):

1. On object mask. I am trying to find a way to improve mask accuracy on edges. Model itself is limited on lower resolution of image encoding and mask generation, so I am looking on reasonable way to improving it during post-processing step using full-res image.
2. On RAW denoise model. It's current architecture is challenging for some GPU execution providers to compile graph effectively. But it works pretty fast even on CPU. Quick fix is just to force CPU execution on problematic EPs. More sustainable fix will require small change in the model itself. I am discussing it with Benoit Brummer.

[wpferguson]

I have one more addition to the Lua API, #20968.

I also have a question about the lua-scripts. Since they are now bundled with the release do they need to "freeze" with the release? At this point the only changes I still want to do are adding the ability to use internal metadata for the scripts that currently spawn exiv2 or exiftool to get metadata. The only other changes would be adding any AI example scripts for #20716. I'll plan on doing another submodule update before the freeze date.

[kofa73]

I don't have anything planned for 5.6. It's too late for the white balance rework to be reviewed, tested and included. Maybe for 5.8.

[dtorop]

@TurboGit: Thank you for pointing out!

The luma waveform/histogram in #20704 could merge into 5.6. But 5.6 already has a bunch of improvements to scopes, and I'm not sure if the PR is resolved (from UI POV).

Besides that, I've got a PR in progress to cajole a bit more code into GTK4-style event handling. The biggest change is for culling, (particularly to make it no longer depend on root window coordinates). While it's a legit code improvement, and will help clear way for a GTK4 migration, I don't think it provides any noticeable improvement for GTK3, and there's always the risk that it subtly breaks something. Regardless, it's not exactly a "feature".

Also: @kofa73 pointed out what looks like a regression. I haven't looked into this yet, but this wouldn't be a feature but a bugfix. My guess is the work to making scrolling consistent (#20843) will spin out into 5.8.

[TurboGit]

We are entering feature freeze. A PR that I still need to commit is #20888, I'll try to test and merge later today.

[kmilos]

I don't think that folks proactively choose HEIF, it's just what they are left with sometimes with phones

The import facility for HEIF is provided precisely so people can get to their important images and memories in the case when they didn't have, or weren't aware of a choice in this. Export is, however, a very much conscious decision IMHO.

IMHO if the legal issue needs more consideration we should not push this feature for the upcoming release last minute and have a proper discussion first on how to address the doubts and the opinions of all core maintainers.

Yep, the whole point of this.

[MStraeten]

regarding zoom: there's a critical regression which disables a zoom by scroll gesture on apple magic mouse

[da-phil]

regarding zoom: there's a critical regression which disables a zoom by scroll gesture on apple magic mouse

Given that you can restore the previous behaviour within the settings, it's not a regression, it's just a lack of support for the Apple Magic Mouse for the new feature.
Let's work on it within the following PR:
#21058

[MStraeten]

if the default behaviour breaks an existing functionality then it’s a regression. Disabling the default functionality is a workaround ;)
That why i added the mac related note to the release readme.

[da-phil]

That why i added the mac related note to the release readme.

Excellent, thanks 😉

[jenshannoschwalm]

Just pinging here.
There is something strange going on in generic tiling code - indeed might be a very old thing. Will investigate the next days. Could possibly be release-critical.

[TurboGit]

@jenshannoschwalm : Noted.