|
From: <no...@tc...> - 2025-10-12 16:03:58
|
Automated mail by fx, on behalf of no...@tc... Ticket Change [fa3d5994df5de5b6219871fd0e63e26d03735ea3f2b300ff969f5843dd8faa7d] [Tests event-9.15 and event-9.20 are broken, and fail on macOS/aqua when fixed] By erikleunissen For Tk On 2025-10-12T16:02:50.032 Details https://core.tcl-lang.org/tk/tinfo?name=fa3d5994df5de5b6219871fd0e63e26d03735ea3f2b300ff969f5843dd8faa7d Ticket https://core.tcl-lang.org/tk/tktview/c494cc25c4afef0425273eee288a7e7b02d260e7 Changed Fields icomment: Analysis and conclusion of the experiments regarding tests event-9.15 and event-9.20 ---------------------------------------------------------------------- -------------- The experiment targetted the following two questions regarding an extraneous <Enter> event on the root window, for tests event-9.15 and event-9.20: A. Reproducibility of the issue, on Catalina, Sequoia and Tahoe B. Origin/cause of the extraneous <Enter> event. Results ------- The raw results are here: [https://github.com/tcltk/tk/actions?query=branch%3Abug-c494cc25c4] Please see as separate files the outputs for test file event.test under macOS Catalina, Sequoia, Tahoe and Linux/x11 (debug and optimized builds distinguished) in the archive file c494cc25c4-experiment.zip. Analysis and conclusion ----------------------- A. Reproducibility of the issue, on Catalina, Sequoia and Tahoe * The failures of tests event-9.15 and event-9.20 have been shown to be 100 % reproducible for macOS Catalina and macOS Sequoia, using machines at Github and machines owned by me. * The test failures haven't occurred a single time on macOS Tahoe (at Github and with 20 runs on the systems used by Marc). This leads me to the most important conclusion for the present ticket: The issue is not to be attributed to Tk, but to the macOS windowing system. The issue exhibits at macOS versions Sequoia and older, and appears to be fixed on macOS Tahoe. B. Origin/cause of the extraneous <Enter> event Given the conclusion at A., the quest for the origin is not a priority anymore. Nevertheless, this ticket raised a hypothesis as an explanation of the extraneous <Enter> event: an overloaded WindowServer process. Regarding that hypothesis, I'd like to point out the following: * the extraneous event only occurs for tests event-9.15 and event-9.20, and at ever the same position in a sequence of window events. * overloading implies late event notifications. However, - the extraneous <Enter> event is not just late: - it occurs extra, or extremely ou ...((truncated)) login: erikleunissen ------------------------------------------------------------ See Tcl/Tk development @ http://core.tcl-lang.org/ ------------------------------------------------------------ |