15527 - [PP][IME-X] Japanese characters are not displayed in text input field

Status: VERIFIED FIXED

(Core :: Internationalization, defect, P1)

Description

[Teruko Kobayashi]

In the text field of the form, after you type Japanese characters and hit the
enter to commit the characters, the characters are not displayed.

Steps of reproduce
1. Go to the above URL
2. Go to the input text fields for Name, Address.., under Customer ID
3. Turn on IME by Shift+Space
4. Type Japanese characters
At this time, the Japanese characters are displayed.
5. Hit the enter to commit the characters
Notice that the Japanese characters you typed are dispeared.

Tested 10-01 Linux M10 build.

Comment 1

[Katsuhiko Momoi]

Note the similarities of this bug to Bug 13454, which was
originally filed for MailNews widgets but seems to apply to
input widgets in general.
There is also another bug filed for form, Bug 15310.

What all these bugs have in common is that they share the same
workaround. The end results of not applying the workaround is
a crash (Bug 13454), freeze/crash (Bug 15310), and no input
can be commited/freezing of the input field (Bug 15527), but the
same workaround can be applied to all 3 bugs. That is to input
1 ASCII character, e.g. a space, before turning on the IME for
JPN and inserting cursor point into the input field.

For this particular bug, we should recommend the following
wrokaround for M10 and later.

1. Without turning on kinput2/wnn4, place the cursor point into
   the input filed in question.
2. Now insert 1 ASCII character, e.g. a space.
3. Turn on kinput2 and input Japanese characters - because of
   another bug, under some conditions, you may not see characters
   before the characters are commited.
4. Once you're done with inputting Japanese, turn off kinput2.
5. You will now see comitted string in the input field.

By the way, once you turn on kinput2 and place the cursor inside an
input field withou tfirst inserting an ASCII character, that field
seems to become disabled/frozen for any input since it does not take
any cursor point insertion thereafter even if you turn off input2.
The similar problem happens in Bug 15310.

All these bugs probably share common causes somewhere.

Comment 2

[Katsuhiko Momoi]

By the way, you can delete the extra ASCII character by
backspacing over it without crashing or freezing and then
you will be left with only the JPN input.
Turn off the IME, place teh insertiong point at the beginning of
the JPN string and backspace to delete the extra ASCII character.

Comment 3

[Katsuhiko Momoi]

Correction on one of the above observations.

The following passage:

"By the way, once you turn on kinput2 and place the cursor inside an
input field withou tfirst inserting an ASCII character, that field
seems to become disabled/frozen for any input since it does not take
any cursor point insertion thereafter even if you turn off input2.
The similar problem happens in Bug 15310."

applies to only Bug 15310. In this bug, you can actually input
JPN but you won't see any characters committed and displayed.
Please disregard this particular passage for this bug.

Comment 4

[Hidetoshi Tajima]

The codes to commit the text are currently commented out by #ifdef USE_XIM_NOT
block
in nsGtkEventHandler.cpp, as it has caused more critical problem that all
control
chararcters and BackSpace don't work. A fix to solve both problems together is
to check if kevent.keyCode is 0 or not. If it is left 0, we can assume the event
comes
from XIM to commit the text, if not, we can dispatch it as a keypressed event.

Comment 5

[Frank Tang]

Mark it as M11 P1. I land the keyEvent_19991004_BRANCH yesterday. Now the code is reenable.
Please take a look at the nwe key event spec and make sure non IME stuff do the right thing-
http://www.mozilla.org/editor/key-event-spec.html

Comment 6

[Frank Tang]

should this mark as fixed ?
I think we should change the current
(!kevent.keyCode)
to
(GDK_VoidSymbol == event->keyval)

I can switch on IME in the current code by Shift+space, but I cannot turn OFF IME by using Shift+space.
I could turn it off in my local build if I use the GDK_VoidSymbol line....

Comment 7

[Erik van der Poel]

Toshi, if this bug has been fixed, please mark it RESOLVED FIXED. If it is a
duplicate of some other bug, please mark it RESOLVED DUPLICATE. If it has not
been fixed, and you don't think you can fix it for M11 (deadline is 11/2 at
11:59pm), then please change the Target Milestone to something after M11.

Comment 8

[Teruko Kobayashi]

This still happens in 11-01-08 Linux build.

Comment 9

[Katsuhiko Momoi]

Note that this bug had been fixed and was working on a Linux build from 10/29/99.
But as reported in Bug 17673 today, there seem to be a major regression in IME input
in that input is now disabled in several major areas. We reported a bug for mail composer
related fields but the problem seems more general and applies here also.

Comment 10

[bobj]

Bug 17673 was marked fixed/verified yesterday.
Is this bug now fixed too?

Comment 11

[Teruko Kobayashi]

I verified this in 11-03-08 Linux build.