From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.6) Gecko/20040115 Description of problem: Backspace on Canna clients sometimes recover deleted text instead of cut tail of the input. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Start kinput2 2. Press 'b' 10 times. 3. Press backspace key 5 times. 4. Press 'b' 1 time. 5. Press backspace key 1 time. Actual Results: Recover deleted text. Expected Results: Delete 1 char at tail. Additional info: This bug happen on all Canna clients. kinput2-canna, nvi-canna(RHL9), xemacs, im-canna(http://bonobo.gnome.gr.jp/~nakai).
I can't seem to reproduce with with a stock RHEL 3 WS (Japanese Install with all RHN updates). Maybe I don't understand your instructions to reproduce. Using it's gnome-terminal, and typing the sequence after an "echo" command, and pressing return (which commits the pre-edit sequence), it seems output only five "ã£", which is what I expected the pre-edit buffer to contain.
Hmm, I could reproduce this problem once, but it seems not always, because I can't reproduce it right now.
OK, if I remove ~/.canna, it works correctly. So if Adrian had ~/.canna, you would see the same problem. If user don't have ~/.canna, the kinput2 candidate for 'shou' is 12 line, buf with ~/.canna, which useradd generates, 24 line. The default ~/.canna might have the problem. I think Tagoh also changed it to use kana key binding, right?
(bhuang is still in QA contact...)
It seems to happen if you enable break-into-roman. and if .canna is missing, it doesn't appear so that break-into-roman is disabled by default.
comment 3: Nakai, could you attach a sample .canna file to this bug so other QA testers can easily reproduce? Thanks!
make sure your .canna file includes (setq break-into-roman t) to reproduce this problem.
Sane guys would use /etc/skel/.canna for test instead of my copy.
Fixed in 3.7p1-5.