Description

chrstphrchvz

opened on May 15, 2023

Issue body actions

cpython/Lib/idlelib/autocomplete_w.py

Lines 185 to 190 in 27d8ecd

	# Put widgets in place
	self.autocompletewindow = acw = Toplevel(self.widget)
	# Put it in a position so that it is not seen.
	acw.wm_geometry("+10000+10000")
	# Make it float
	acw.wm_overrideredirect(1)

Using wm geometry to move an overrideredirect toplevel far enough offscreen triggers a strange bug in recent Tk Aqua 8.7 (assuming it isn’t actually a macOS bug) which leaves the toplevel completely blank. I have reported it upstream: https://core.tcl-lang.org/tk/tktview/132dd3d350

However, I would like to see if IDLE can make a simple change to avoid the issue, in case Tk Aqua does not address it soon. Moving a toplevel far offscreen to hide it while configuring its contents seems improper, as opposed to something like withdrawing it. Hiding the toplevel to avoid visual artifacts is probably no longer necessary, except maybe on something sufficiently slow like X11 forwarding. IDLE also does not make a similar effort to temporarily hide the toplevel for calltips.

So I would suggest making this wm_geometry() call only when acw._windowingsystem == 'x11' or at least acw._windowingsystem != 'aqua'.

Linked PRs

• gh-104499: IDLE - fix completions for tk aqua 8.7 #104591
• [3.11] gh-104499: IDLE - fix completions for tk aqua 8.7 (GH-104591) #104596
• gh-104499: Fix typo. #104598
• [3.11] gh-104499: Fix typo. (GH-104598) #104599

Activity

1. terryjreedy

   added

   topic-IDLE

   on May 15, 2023
2. terryjreedy

   added this to IDLE Issueson May 15, 2023
3. terryjreedy commented on May 15, 2023

   

   terryjreedy

   on May 15, 2023

   Member

   More actions

   I always thought moving offscreen instead of withdrawing, as done elsewhere while setting geometry, was a bit weird, but changing was not a priority before now. I think I would use query.py as an example.
4. terryjreedy

   self-assigned this

   on May 15, 2023
5. terryjreedy

   added 2 commits that reference this issue on May 17, 2023

   pythongh-104499: IDLE - fix completions for tk aqua 8.7

   1f57426

   gh-104499: IDLE - fix completions for tk aqua 8.7 (#104591)

   678bf57
6. miss-islington

   added a commit that references this issue on May 18, 2023

   pythongh-104499: IDLE - fix completions for tk aqua 8.7 (pythonGH-104591

   ff8ef9d
7. JelleZijlstra

   added a commit that references this issue on May 18, 2023

   pythongh-104499: IDLE - fix completions for tk aqua 8.7 (python#104591)

   a5a7344
8. terryjreedy

   added 3 commits that reference this issue on May 18, 2023

   pythongh-104499: Fix typo.

   987c304

   [3.11] gh-104499: IDLE - fix completions for tk aqua 8.7 (GH-104591) (#…

   e96fb39

   gh-104499: Fix typo. (#104598)

   c5b670e
9. miss-islington

   added a commit that references this issue on May 18, 2023

   pythongh-104499: Fix typo. (pythonGH-104598)

   2e9481e
10. terryjreedy commented on May 18, 2023

    

    terryjreedy

    on May 18, 2023

    Member

    More actions

    I am a little surprised that there is no withdraw for calltips. It might have been dropped when we redid tooltip and based calltip on tooltip. So I might review the need for withdrawal by checking performance on my 2012 MacBook Air. However, in some cases, tests skip deiconify so as to minimize visual flutter.

    Making windows deiconified for tests seems needed when simulating keypresses and mouse actions. I have wondered whether the latter would still work if the test window were put off screen.
11. terryjreedy

    added a commit that references this issue on May 18, 2023

    [3.11] gh-104499: Fix typo. (GH-104598) (#104599)

    99a162a
12. terryjreedy

    closed this as completedon May 18, 2023
13. github-project-automation

    moved this to Done in IDLE Issueson May 18, 2023

Issue actions