| Ticket UUID: | 3609681 | |||
| Title: | terminal loss of focus involving shortcuts and gitk/wish | |||
| Type: | Bug | Version: | current: 8.6.0 | |
| Submitter: | drizzd | Created on: | 2013-04-01 09:45:57 | |
| Subsystem: | 55. [focus] | Assigned To: | jenglish | |
| Priority: | 5 Medium | Severity: | Minor | |
| Status: | Open | Last Modified: | 2026-09-19 11:19:30 | |
| Resolution: | None | Closed By: | nobody | |
| Closed on: | ||||
| Description: |
I recently reported this to gnome: https://bugzilla.gnome.org/show_bug.cgi?id=695561 . In some circumstances, when wish exits, keyboard events are no longer received/processed in the terminal window from which wish was started. (I described this as a "loss of focus", but I am not sure if window focus is the issue.)
In gnome-shell, I can only reproduce with the wish application gitk. If I try to simplify gitk, the issue appears only sporadically (e.g. once in 10 times).
But in fallback mode (gsettings set org.gnome.desktop.session
session-name 'gnome-fallback') it can be reproduced easily and reliably by doing either one of:
echo 'after idle { exit }' | wish
echo 'bind . <Control-q> { exit }' | wish
I failed to reproduce this with a small program doing XOpenDisplay, XCreateSimpleWIndow, XSelectInput, XMapWindow, XNextEvent, XCloseDisplay. What might Tk be doing differently?
| |||
| User Comments: |
serhiy.storchaka added on 2026-09-19 09:24:38:
Cannot reproduce with current Tk and the ibus XIM server: an xterm keeps receiving keyboard input after wish started from it exits. SCIM is no longer maintained or shipped by distributions. kbk added on 2015-05-26 03:29:01: I vaguely recall reading about input methods way back when 8.4 was getting ready for deployment. KHIM, while it has 'input method' in its name, is merely a set of Tk keybindings, and has no connection with X input methods. I really don't understand them, except that to use them successfully appears to depend on several client libraries that we don't link to. The most approachable discussion of how they work appears to be http://menehune.opt.wfu.edu/Kokua/Irix_6.5.21_doc_cd/usr/share/Insight/library/SGI_bookshelves/SGI_Developer/books/XLib_PG/sgi_html/ch11.html - and even at that, I have a hard time making sense of it. I vaguely recall Joe discovering something in how we were handling keyboard events that accounted for some odd behaviours under SCIM. I don't even recall what the outcome of the discussion was. Kicking it over to Joe in hopes he remembers. aku added on 2015-05-25 16:49:10: Back to Kevin. anonymous added on 2015-05-25 03:56:35: Mention of SCIM identifies this ticket with [3610493fff]. ferrieux added on 2013-04-01 17:55:30: Pinging Kevin, I think input methods are in his (numerous) areas of expertise. drizzd added on 2013-04-01 17:47:17: Here is one such message:
0.01: Client --> 24 bytes
............REQUEST: ConvertSelection
requestor: WIN 04c00001
selection: <@server=SCIM>
target: <LOCALES>
property: <LOCALES>
time: CurrentTime
0.01: 32 bytes <-- X11 Server
..............EVENT: **INVALID** (159)
time: CurrentTime
requestor: WIN 04c00001
selection: <@server=SCIM>
target: <LOCALES>
property: <LOCALES>
drizzd added on 2013-04-01 17:45:40: Indeed there are some suspicious events which appear to be related to scim (my X input method daemon). If I turn off scim in the calling terminal (unset XMODIFIERS; xterm), the problem goes away. I have attacked xscope output both with and without scim. drizzd added on 2013-04-01 17:41:34: File Added - 462152: xscope-noscim.out drizzd added on 2013-04-01 17:41:05: File Added - 462151: xscope.out ferrieux added on 2013-04-01 16:58:00: Try inserting xscope between wish and the X server; with luck one of the X events or commands will stand out. | |||
Attachments:
- xscope-noscim.out [download] added by drizzd on 2013-04-01 17:41:34. [details]
- xscope.out [download] added by drizzd on 2013-04-01 17:41:05. [details]
