Backspace don't work with voweled Arabic text

[khaledhosny]

Try a simple string like نَ, 'backspace' can't delete it, but 'delete' can. It looks like the mark get reinserted immediately after being deleted. Using svn revision 11359.

[friedelwolff]

This has to do with combining unicode characters. The same will happen with ê (e + combining circumflex).

[friedelwolff]

My first guess is that this is because of much of the new code dealing with iteration positions in the buffer, and since these are non-spacing characters, they might be messing up our calculations. I guess this is pretty serious for Indic and almost all South East Asian languages. For Arabic this mostly affects transliterations, or am I wrong?

[khaledhosny]

(In reply to BZ-IMPORT::comment #2)

I guess this is pretty serious for
Indic and almost all South East Asian languages. For Arabic this mostly affects
transliterations, or am I wrong?

No, that is serous bug of Arabic too. Vowel marks are integral part of Arabic script and are generally used to disambiguate ambiguous words which is not limited to transliteration, though fully vocalized text is uncommon.

[friedelwolff]

This was reported by leuce in IRC, and sounds as if it might be related:

Target text:
Kies een 256x128 PNG formaat afbeelding voor het label.

If you go to that segment, it turns the TL into
Kies een 256x128128128 PNG formaat afbeelding voor het label.

If you go down and up again, it turns the TL into
Kies een 256x128128128128128 PNG formaat afbeelding voor het label

In the same segment as #2: put your cursor just after the "x", and hold in Del. Virtaal deletes some numbers but then reinserts them again, over and over.

The middle issue is probably something separate, and is tracked in bug #1016, but the last part sounds as if it might be similar to the issue in this bug.

[walterl]

Reported the (py)gtk bug that causes this problem at http://bugzilla.gnome.org/show_bug.cgi?id=588089

[walterl]

Fixed in revision 11912... Finally! :D

It does confuse undo a bit, but that is covered in bug #1048.

The PyGTK bug I filed seems to be more of a symptom of the real bug, but I'll report the GTK bug where appropriate.