From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4) Gecko/20030703 Description of problem: kinput2 mysteriously becomes inoperative if you click on a character choice within the Candidate Selection dialog after you have dragged that dialog window. If you never click & drag the Candidate Selection dialog, it works fine. Unfortunately due to the annoying problems of gtk2 apps like Bug 84860 and Bug 84859 where the candidate selection window appears below the bottom left of the window rather than under the text input, it is guaranteed that the user will click & drag the kinput2 Candidate Selection dialog, triggering this bug. Version-Release number of selected component (if applicable): kinput2-canna-wnn6-v3.1-10 How reproducible: Always Steps to Reproduce: 1. SHIFT-SPACE 2. nihongo 3. SPACE SPACE 4. Click and move Candidate Selection window anywhere. 5. Click on 日本語 (the far left choice) Actual Results: kinput2 seems to no longer respond. Western character input continues to work in gedit, but within Mozilla keyboard input behaves strangely and unable to enter western characters. Expected Results: It should not render kinput2 inoperative.
It looks like similar bug with Bug#78017. and I can't reproduce this except gtk2 applications.
Hmm, this sometimes causes kinput2 to fail totally and consume 100% CPU. Sometimes it fails completely only after several tries only in that individual application, and subsequently works again if you restart that application (like gedit).
When it got stuck at 100% CPU usage, strace kept repeating this in a seemingly infinite loop: --- SIGPIPE (Broken pipe) @ 0 (0) --- select(6, [4 5], [], [], {0, 0}) = -1 EBADF (Bad file descriptor) getuid32() = 500 geteuid32() = 500 getuid32( kill failed to kill it. It took kill -9 to stop it.
I suspect it's maybe another bug but it's just triggered by this. Could you please submit it as another bug and mention the sure way to reproduce the problem?
Yes, I think the 100% CPU usage and kill -9 problem is a separate bug. Maybe Bug #114886.
This is no longer imperative for FC3/RHEL4 as IIIMF is the preferred input method. Please verify that this is not an issue for RHEL3 customers.
WONTFIX in RHEL3 Updates.