Merge pull request #850 from Psychtoolbox-3/master #279
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Psychtoolbox 3.0.19 update "Last free lunch"
Compatibility changes wrt. Psychtoolbox 3.0.19.14:
Highlights:
All:
Change enumeration order of video capture devices. This should avoid selection of suboptimal capture devices by preferring more optimal ones, especially on modern Ubuntu 22.04-LTS and later Linux distributions, which often choose the new but immature pipewiresrc as video source plugin for a camera instead of the old and proven/mature more optimal v4l2src as video source. Specifically we hope to avoid VideoRecordingDemo.m problems like the ones encountered in the following thread:
https://psychtoolbox.discourse.group/t/codecs-in-videorecordingdemo-not-working-on-linux/5441
PsychImaging: Fix HDR mode with types of PsychColorCorrection() that involve lookup tables and only use one color correction operation in total. E.g, standard LUT based correction, 3D LUT based correction or application of per-pixel gain matrices. See:
https://psychtoolbox.discourse.group/t/best-approach-for-gamma-correction-in-hdr10/5515
CI release scripts: Disable building of .mltbx files for releases for now.
The current approach contributed by Mathworks team is broken. It includes the whole Psychtoolbox-3 tree for some utterly unclear reason, although the project file only specified the Psychtoolbox/ subfolder. Worse, it adds toolbox folders to the Matlab path in randomized (!?!) order, so our carefully selected ordering of MatlabWindowsFilesR2007a/ folder wrt. PsychBasic folder gets screwed up and mex files can't be found on MS-Windows anymore. Even worse, if SetupPsychtoolbox() fixes it, then the Matlab Add-On Manager undos the fix every single time Matlab restarts! This leads to issues like the following linked one:
https://psychtoolbox.discourse.group/t/problem-with-screen-function-ptb-just-installed/5451/9
Apart from that, the Add On manager doesn't allow toolbox updates, and doesn't set things like the Matlab Java path, doesn't deal with GStreamer etc. It has zero benefit for our users, but is a source of hazarads and extra support overhead. Drop it.
Add Psychtoolbox/PsychBasic/PsychPlugins/ to the search path for dynamic shared libraries for all operating systems, so PTB specific plugin libraries can be found during mex file load time.
Linux:
Psychtoolbox was built and tested against Matlab R2024a and Octave 5.2 and later.
PsychImaging: Update 'EnableNative16BitFramebuffer' help text for year 2024.
Screen: Fix Vulkan display backend support with Mesa versions 23.3 and later and thereby for Ubuntu 24.04-LTS and later. The same problem that affected the Broadcom VideoCore 6 of the RaspberryPi 4 and later under Mesa 23.3 now affects at least Intel and AMD under Mesa 24.0 with Ubuntu 24.04.0-LTS.
It appears a slightly broken GL_EXT_direct_state_access implementation in Mesa was introduced in Mesa 23.3 and later. Work around this in an effective and safe way.
PsychVulkan: Add workaround against an AMDVLK bug that causes crash if the used Vulkan display monitor has more than 64 video modes. A proper fix, especially for multi-display setups, can only be done in a future AMDVLK driver from AMD's Vulkan team, but for single display stimulation this workaround should prevent crashes and problems, and on some multi-display setups things may still work with a sprinkle of dumb luck - no guarantees for multi-display though! Note that the AMDVLK driver is only used on AMD gpu's if a color depth of more than standard 8 bpc (aka "deep color") or HDR-10 display mode is requested. For other Vulkan use cases, the high quality Mesa radv Vulkan driver is used, which does not suffer such a problem. For context see:
https://psychtoolbox.discourse.group/t/vulkan-breaks-with-stack-smashing-detected-terminated/5501
XOrgConfCreator: Adapt to Ubuntu 24.04-LTS changes in /dev/dri/card enumeration. This broke hybrid graphics laptop detection, at least on three single gpu laptops with AMD gpu's. It reported a dual-gpu config when there was not any. Since Ubuntu 24.04-LTS, or more likely with a recent Linux 6.8 kernel shipping in both Ubuntu 24.04-LTS and in Ubuntu 22.04.5-LTS with HWE stack, the first (and only) gpu on a single gpu machine now has a /dev/dri/card1 device file, instead of a /dev/dri/card0 device file, ie. device file indexing now starts at 1 instead of 0. On multi-gpu machines, the second gpu is now associated with /dev/dri/card2 and so on. This broke some hard-coded assumptions about 0-based device numbering. We now use a more robust detection logic for hybrid gpu detection. The fix was successfully tested on single-gpu AMD and Intel graphics machines, and on a dual-gpu Intel + AMD laptop for each gpu and in prime render offload mode.
Add braces for NetWM reporting to make compiler happier. It caused a weird problem in the Screen mex build for Matlab, in which the "PTB-INFO: ..." lines about the used flip timestamping mechanism are not printed! No clue why, looks like some compiler bug, but adding the - technically not needed - braces fixes it.
Windows:
Psychtoolbox was built and lightly tested against Matlab R2024a and Octave 7.3.
VideoRecordingDemo.m: Remove special case MS-Windows path. It does not seem to be as necessary anymore with the future default MediaFoundation path for capture and recording. Also added some cleanups and ability to specify video capture device as argument.
macOS: