Consider the following situation.
Affix file:
SET ISO8859-1
COMPOUNDBEGIN x
COMPOUNDEND z
COMPOUNDPERMITFLAG c
ONLYINCOMPOUND oSFX j Y 1
SFX j 0 0/xc .
Dictinary file:
2
Arbeits/j tt:test1
fundament/z tt:test2
In Hunspell 1.3.3 the result of analyze is:
analyze(Arbeitsfundament) = pa:fundament st:fundament tt:test2
But it should be:
analyze(Arbeitsfundament) = pa:Arbeits st:Arbeits tt:test1 fl:j pa:fundament st:fundament tt:test2
In spell checking mode, Hunspell correctly recognizes “Arbeitsfundament” as correctly written word. So this bug affects only the analyze mode, but not the spell checking mode.
If you remove the flag “o” of the last line in the affix file
then you get the correct analyze result.
Alternatively, you can also modify the dictionary file
and you get correctly
By the way: This is a real-world example. The data is an extract of the german de_DE_frami dictionary.