Created attachment 103972 [details] Core dump from the crash On OSX 10.9.4 / LO 4.3.0.4 release: Steps to reproduce: 1. Create a new Writer document 2. Insert a comment 3. Within the comment, start typing any Japanese text with the 'Kotoeri' - 'Hiragana' input method. For instance, 'aaaaaaaaaaaaaaaaa' (which will produce the text 'ああああああああああああああ') 4. Without confirming the input conversion (i.e. without pressing 'Return'), select a new font from the font dropdown 5. Either a crash will occur at this point (see attached backtrace), or a hang when the next Japanese text is input This appears to be specific to comments - I have not observed the same crash occuring in the body of the document
Additional note: When the above results in a hang rather than a crash (i.e. when the font selection is initially successful rather than crashing immediately), I notice that the text that is being entered is deselected, indicating that the input conversion has been aborted When doing the same thing in the main document text, the text remains selected, and the conversion is not aborted, so the font can be changed in mid entry. This should also be possible in a comment * I've just spotted a coincidental but probably(?) unrelated issue; in the main document, when changing the font while a Japanese input conversion is open, the font is correctly applied to the whole text of the input, but *only when the next character is entered*. Raised separately as bug 82116
Hi Matthew, Thanks for this new bug and crash report. Adding the Mac QA team to confirm.
Hi, to reproduce I'd need to know how to "start typing any Japanese text with the 'Kotoeri' - 'Hiragana' input method". Sorry I've not once in my life used those method - whatever they may be.
(In reply to comment #3) > Hi, to reproduce I'd need to know how to "start typing any Japanese text > with the 'Kotoeri' - 'Hiragana' input method". Sorry I've not once in my > life used those method - whatever they may be. In OSX 10.9 it's like this: (earlier OSX may be slightly different) 1. Open "System Preferences" from the Apple menu 2. Select "Keyboard" 3. Select the "Input Sources" tab at the top of "Keyboard" 4. Click on the "+" button at the bottom left to add a new input source 5. Select "Japanese" on the left, then "Kotoeri" on the right and click on "Add" 6. Back in the "Keyboard" - "Input Sources" tab, make sure that at least "Input modes:" "Hiragana" is enabled in the Kotoeri options (this should be enabled by default) 7. At the bottom of "Keyboard" - "Input Sources", make sure that "Show Input menu in menu bar" is enabled (again, probably enabled by default) To test: 1. Open a Writer document 2. With the writer document focused, open the input menu, which is an icon somewhere on the right side of the system menu bar, and select the option that looks like "あ Hiragana" 3. Type some "aaa"s, which should appear as "あああ" At this point, you should be pre-editing some Japanese. What would normally happen next is that you either press "Space" to convert the phonetic text you've entered (a popup will appear at the cursor with some options), and/or "Return" to accept the text as it is currently. While in the pre-editing mode, the text at the cursor that the input method is handling is underlined, and in the case of LO also appears as selected text in the document Reproducing the bug involves changing the font while the pre-editing mode is open within a comment, before pressing "Return" to confirm/exit.
Confirmed:OSX:4.3.0.4 -> NEW Thanks Matthew for making this idiot proof description (Comment 4, and then back to initial Description). I was able to reproduce the crash. Attaching a video (not too much too see - instead of mouse pointer, I was seeing the psycho pizza of death, after changing the font to "Bauhaus 93"). Interestingly the crash did not occur, when I changed the font to Times New Roman. Crash log not attached, since we have that from Matthew already.
Created attachment 104410 [details] video
It looks like the originally attached backtrace actually related to the now fixed bug 82260 - the traces seem identical. The issue that Foss successfully reproduced is however still present in git (which I have now built), and under a development build gives a rather different backtrace, which I will attach.
Created attachment 104421 [details] Updated backtrace
Occurs all the way back to 3.3.0 (and still in 4.4.0.2) Setting Version to "Inherited from OOo"
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.