Crash on adding new word to dictionary

[Madarb]

Aegisub produces a crash when I try to add a specific word to the dictionary. The word is Führer and I suppose the problem is met because of the "ü" character, because Aegisub never failed adding new words under any other circumstances.

Version is r3177M (development version, TheFluff).

[TheFluff]

Works here with r3542 on Windows XP, US English dictionary. What language are you trying to add it to?

[Madarb]

Greek.
Just tried it on the stable 2.1.7 [3131] and it does the same.
Reproduce steps:

1.Open subtitle file with the faulty word.
2.Click the spell checker.
3.Find the highlighted word, and select "Add to Dictionary".
4.Kaboom.

[verm]

We've got so many languages now it's probably best we have this fixed for 2.1.8.

[Madarb]

Requested greek dictionaries

[verm]

I can confirm this is defiantly broken in more ways than one.

• The 'Add XXX to dictionary' menu item is never highlighted
• Adding the word via 'SpellCheck this script' -> add word crashes aegisub.

I converted the dictionaries from ISO-8859-7 to UTF-8 and deleted my personal dictionary before testing.

[Plorkyeran]

Cannot reproduce. Steps followed:

1. Deleted AppData/Aegisub
2. Installed fresh copy of 2.1.7
3. Added Greek dictionary to AppData/Aegisub/dictionaries
4. Attempted to add Führer; menu item was correctly greyed out
5. Converted Greek dictionary to UTF-8 with iconv
6. Successfully added Führer via the context menu

[Plorkyeran]

Also works correctly for me on XP with the non-unicode locale set to Greek.

[Madarb]

Um, I wasn't aware of any converting that needed take place. :/
Does this work via SpellChecker too?

Don't know if it matters, I only tried it under Vista here...

[Plorkyeran]

ü doesn't exist in ISO-8859-7, so it's impossible to store the word Führer without converting the dictionary to a charset that does have both ü and Greek letters. I'll attach the UTF-8 version of the Greek dictionary, which may fix your problem.

There's a two ways I can think of to resolve this:

1. Instead of simply not allowing invalid words, ask the user if they want to convert their dictionaries to UTF-8 (i.e. ask the user a question they probably won't know the answer to).
2. Always save Aegisub-generated dictionary files as UTF-8 and modify the spellchecker code to handle different loaded dictionary files having different character sets.

Neither of these seem reasonable for 2.1.8.

I suspect the actual crash is something similar to #1108 and #867.

[verm]

Interesting it does work for me now I must have made a mistake last night and didn't realise.

Now we're onto the next problem: refusing to load a dictionary if it's not UTF-8.

[plorkyeran]

(In [3994]) Disable the "Add to dictionary" button in the spell checker dialog when the word cannot be added to the dictionary. Updates #999.

[plorkyeran]

(In [3995]) Merge r3994 from the 2.1.8 branch. Closes #999.