GNU bug report logs - #71912
30.0.50; Inaccurate window-absolute-pixel-position Values on macOS After Sleep and During Fullscreen Mode

Previous Next

Full log

Message #8 received at 71912 <at> debbugs.gnu.org (full text, mbox):

From: Gerd Möllmann <gerd.moellmann <at> gmail.com>
To: Jones Stephen <happyojones <at> gmail.com>
Cc: 71912 <at> debbugs.gnu.org
Subject: Re: bug#71912: 30.0.50; Inaccurate window-absolute-pixel-position
 Values on macOS After Sleep and During Fullscreen Mode
Date: Tue, 23 Jul 2024 13:06:09 +0200

Jones Stephen <happyojones <at> gmail.com> writes:

> Dear Emacs Developers,
>
> I am encountering an issue with window-absolute-pixel-position on macOS that leads to incorrect Y-coordinate values under specific conditions. This
> bug appears to manifest in two scenarios: after the system resumes from sleep, and when Emacs is in fullscreen mode.
>
> Environment:
>
> Emacs Version: (30.0.50)
> Operating System: macOS (After 14.00)
> Issue Description:
>
> After Sleep: When macOS resumes from sleep, the Y-coordinate value returned by window-absolute-pixel-position is off by approximately one-fourth
> of the screen height.
> During Fullscreen Mode: In fullscreen mode, the Y-coordinate is incorrectly offset by the height of the navbar.
> Steps to Reproduce:
>
> Open Emacs and position a window at a known pixel position.
> Either put the system to sleep and wake it, or switch Emacs to fullscreen mode.
> Evaluate (window-absolute-pixel-position) and note the Y-coordinate.
> Expected Behavior:
> The function should return accurate window positions regardless of sleep mode or fullscreen state, without any unexpected offsets.
>
> Actual Behavior:
> The Y-coordinate is inaccurately calculated, showing a significant offset that seems to be influenced by system sleep and fullscreen transitions.
>
> This issue can be particularly disruptive in workflows that rely on precise window positioning, such as automated GUI testing or window management
> scripts. Any insights or fixes would be greatly appreciated!
>
> Thank you for your attention to this matter.
>
> Best regards,
> happyo

I can reproduce this (only tried the fullscreen case), but I don't know
how to fix it. In a fullscreen frame, eval

  (ns-frame-edges (selected-frame) 'native-edges)
  -> (566 39 2290 1107)

That is, both left and top are bogus values. Also,

  (frame-parameter (selected-frame) 'fullscreen)
  -> fullboth
  (frame-parameter (selected-frame) 'top)
  -> 39
  (frame-parameter (selected-frame) 'left)
  -> 566

which is why ns-frame-edges in its current implementation returns these
values for left and top. Don't know what is expected here. Should top
and left frame parameters have been set to 0 somewhere? Or is the
implementation of ns-frame-edges wrong in using top_pos and left_pos of
the frame?

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.