https://bugs.kde.org/show_bug.cgi?id=524881
--- Comment #3 from [email protected] --- Additional evidence from Fedora 44 KDE/Wayland, KPipeWire/Plasma 6.7.5-1.fc44, PipeWire 1.6.9-1.fc44, Qt 6.11.2 and an NVIDIA RTX 4080 SUPER: In the retained journal sample through 2026-10-05, I found 462 failed video-link negotiations. The affected consumer advertised RGB, RGBA, xBGR, ABGR, RGBx, BGR and GRAY8, while KWin offered BGRA/BGRx. In the October 3-4 sample, 459 failures belong to one continuing plasmashell process. These are negotiation counts, not audio dropouts or crashes. I tested the extracted, unmodified 6.7.5 buildFormat() with a modeled onRenegotiate offering loop and the observed 13-modifier case. Nine formats with and without modifiers require 4,896 bytes at the default frame-rate setting. With the original 4,096-byte builder, the failing format order yields only 14 valid PODs, excluding BGRA/BGRx. With the 16,384-byte capacity from MR 283, it yields all 18. With a 143 fps cap, the original yields 13 and the patched version 18. Order/capacity checks also preserved guard bytes. This supports a format-list truncation explanation for the process-dependent behavior. I also backported MR 283's capacity/check changes to the Fedora 6.7.5 source RPM, retaining Fedora's existing patch, and built the complete package in a separate Fedora 44 container. The unchanged upstream appstreamtest, TestEncoder and mediamonitortest passed (3/3). Runtime dependencies, library SONAMEs and strong exported symbols were unchanged. Hardware VAAPI and unavailable x264 cases were not covered. The local RPM has not been installed, so this is builder/build evidence, not an on-desktop verification of the fix. The 16 KiB builder is still finite: a 200-modifier stress case can still exceed it and produce a partial list with a warning. Relevant MR: https://invent.kde.org/plasma/kpipewire/-/merge_requests/283 Please let me know whether a compact standalone builder test would help review the fix or a maintenance-branch backport. -- You are receiving this mail because: You are watching all bug changes.
