210134 - JA-IME: can not commit the non-first candidate un-commit character by hit enter on linux RH8.0

Status: RESOLVED WONTFIX

(Core :: Widget: Gtk, defect)

Description

[Yuying Long]

Build: 06-20 1.4 branch build on linux RH 8.0

This problem is on linux RH8.0, but not on RH7.3.  And it is not a regression
from N7.02 (by the time we ship N7.02, I didn't have RH8.0 only RH7.3).

Steps to reproduce:
1. Launch browser, open Composer or any other input fields.
2. Press Shift_space to invoke the Japanese IME.
3. Start type some thing which can bring up more than one candidate characters,
e.g. kannji.
4. Hit space key to bring up the candidate window.
5. Press space to select anything other than first one choice.
6. Press enter to commit.

Result:
Can not commit. 

7. No matter how many times you hit the enter, the string won't change.

The work around is you move the mouse focus to the compose window, hit enter,
then will commit.

This is very bad especially when you try to input a long string.  And I don't
see it on gedit.

Comment 1

[bobj]

CC'd ftang and momoi

Comment 2

[Katsuhiko Momoi]

Can you type the number corresponding to the choice you want
and then hit enter. Is that possible?

Comment 3

[Yuying Long]

Do I understand you correct? there is no number associated with the candidate.

Comment 4

[Yuying Long]

Xianglan found out it's not in mozilla 1.0.1.  And I saw it on N7.02 (1.02).  So
looks like mozilla problem.

Comment 5

[Masaki Katakai]

I think the exact problem is in metacity. RH8's metacity has the problem.
When I updated this to the latest metcity, this problem did not happen.

For RH9, metacity has been updated and does not have the problem.
So there is no problem for Mozilla.

There are two options for RH8,

1. Install new metacity
2. Use over-the-spot input mode

I have no idea to modify our Mozilla to solve this problem.

Comment 6

[Katsuhiko Momoi]

Thanks to Katakai-san we now have critical additional info on
the default Metacity Windows Manager on RH8 & RH9.

For RH8 users using "on-the-spot" default, either they can change the option
to "over-the-spot" or update to the new version of Metacity Windows Manager.

To change the option to "over-the-spot", see this document:

http://www.mozilla.org/projects/intl/m14ximcheckins.html

user_pref("xim.input_style", "over-the-spot")

should be able to avoid this problem.

To summarize:

RH7 and RH 9 users should not have this problem.
RH8 users can take one of 2 ways to avoid the problem as described above.

I suppose we can release note this rather than ship with the 
"over-the-spot" default for all versions of RH Linux. What do you 
think?

Comment 7

[Katsuhiko Momoi]

For metacity versions in various Linux distributions, see this
page:

http://rpmfind.net/linux/rpm2html/search.php?query=metacity&submit=Search+...

Comment 8

[Samir Gehani]

adt: nsbeta1+/adt3

Comment 9

[Katsuhiko Momoi]

The choice is to either rel note this for RH8 users or
reset the default input style to "over-the-stop". If the latter,
we need to make a change to unix.js probably.

Comment 10

[Masaki Katakai]

Writing this to rel notes is the best solution, I think.

Comment 11

[Warren Togami]

> Xianglan found out it's not in mozilla 1.0.1.  And I saw it on N7.02 (1.02).  
> So looks like mozilla problem.

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=62307
mozilla-1.0.1 in RH8 had the over-the-spot prefs.js option added by default due
to my Red Hat Bugzilla report here.  In RH9 that option disappeared, but it
didn't seem to matter because that option didn't seem to work in later 1.2.1 and
1.4 versions of Mozilla in my testing.

I have been fighting Japanese input problems (mainly the one described in Bug
143540) for several years now, and I have never experienced the behavior
mentioned at the top of this report.

https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=84859
Unresolved mozilla-1.2.1 and mozilla-1.4 kinput2 window placement problem in RH9
and higher.

Comment 12

[Warren Togami]

Another issue which really confuses me when everyone (Red Hat, you folks)
mention on-the-spot being preferable, I have NEVER been able to get on-the-top
to work in Red Hat Linux Mozilla.

Comment 13

[Katsuhiko Momoi]

This has been release noted for 1.4.

Comment 14

[Jungshik Shin]

For now, assign to myself

Comment 15

[Masayuki Nakano [:masayuki] (he/him)(JST, +0900)]

per comment 5 and comment 6, this bug is out-of-date.