PreviousNextTracker indexSee it online !

(311/314) 3601097 - SpellCheck incorrectly highligts unknown words

I'm experiencing difficulties using SpellCheck plugin (R007) + Aspell (0.50.3) on Windows (7); JEdit version being 5.0.0. I check a set of DocBook XML files, encoded in Windows-1250, Czech language, indented.
However a significant portion of the words is not highlighted correctly when a Spell check dialog has appeared about what to do with an unknown word. The number of highlighted letters seems to be correct, but the beginning and the end of the highlighted area is incorrectly located on the line. From there on attempts to replace the word end up in weird text combinations.
When I tested spell checking with the same text on Linux with Aspell 0.60.6, I didn't notice the problems. I tried to use Aspell 0.60.6 exe compiled by the Bluefish project, but the exe file crashed every time it was instructed with parameters to use its SGML mode. Then there are Aspell DLLs created by the LyX project and by the UltraEdit project, but these (unlike hunspell.dll) cannot be invoked directly from the SpellCheck plugin -- and I didn't manage to discover any workaround nor did I manage to compile Aspell myself from the source code to a windows exe file.
Neither using Hunspell is a solution because its HTML mode checks the content of the attributes inside the XML tags and thus it reports way too many useless instances of unknown "words".
So is it possible to do something about the Spell Check plugin so that it can be used for adequate spell checking on the Windows platform?

Submitted profant - 2013-01-16 - 12:03:47z Assigned kerik-sf
Priority 5 Category None
Status Pending Group None
Resolution Fixed Visibility No


2013-01-16 - 14:40:51z
you should persist in using hunspell.
To prevent it from checking the content of attributes,
- go to Plugin Options > Spell Check > Syntax Handling
- edit the markup mode to uncheck the LITERAL1 tokens
- OK, OK
Does it work for you this way ?
2013-01-16 - 16:42:55z
This only works (and actually worked, I just didn't notice exactly) in some cases. For example tag <emphasis role="bold"> was spell-checked with LITERAL1 checked ("bold" isn't a Czech word) and without it isn't spell checked.
But tags like <xref linkend="wom_um_uvod">, here "uvod" is marked as unknown. Or <xref linkend="wo_spr_uzivmista">, "spr" is marked even without the LITERAL1 checked,
2013-04-04 - 20:17:58z
fixed problem with unknown words middle of attribute when LITERAL1 is unchecked in;a=commit;h=ff3a4eb4c4487a1efaa9e946569d92845c4c8473
It will be part of release R008 of SpellCheck, this month.
2013-05-01 - 09:43:00z
R008 released April 23, 2013
2013-05-07 - 08:45:38z
R008 didn't fix it properly.
Latest change (4efcca0) really fixed it.
you can get it from