bug#19558: ispell in 24.4: hunspell cannot check German text

[Eli Zaretskii]

> Date: Sat, 10 Jan 2015 15:02:53 +0100
> From: Heinz Rommerskirchen <heinz@h-rommerskirchen.de>
> 
> When using hunspell ispell cannot check German texts.
> 
> To recreate the problem create a file containing only the one line
>  > zwanzigjährigen Arbeitszeit bei der Motorenfabrik Jank gebracht.
> (This is a valid fragment of German, but 'Jank' is a proper name unknown
> to hunspell). I have used both latin-0 and utf8 encoding with no visible
> difference.
> $ emacs -Q
> in *scratch buffer execute (setq ispell-program-name "hunspell")
> open the file and type M-x ispell.
> I got the error message
> ispell-process-line: Ispell misalignment: word `Jank' point 52; probably
> incompatible versions
> 
> If you delete the first word in the file, ispell works fine and flags
> the unknown 'Jank'.
> 
> hunspell on the command line has no problems with this files.
> I was also able to use hunspell by adding the following to my start files
> (add-to-list 'ispell-local-dictionary-alist '("deutsch" "[[:alpha:]]" 
> "[^[:alpha:]]" "[']" t ("-d" "de_DE") nil iso-8859-1))
> 
> 'hunspell --version' gives
>  > $ hunspell --version
>  > @(#) International Ispell Version 3.2.06 (but really Hunspell 1.3.2)

If this is an unpatched version 1.3.2 of Hunspell, then that's
probably a known problem with Hunspell: it reports byte offsets of
misspelled words rather than character offsets, something that Emacs
doesn't expect.

There are patches in the Hunspell bug tracker to fix this problem.

FWIW, your test case works flawlessly for me in Emacs 24.4 with
Hunspell 1.3.2 patched to fix that problem (and a few others).