It seems ispell converts some characters to other groups of characters.
So, a ņ (as in espaņol) is actually stored as ESPA'NOL. There seems to be
a bit of code which converts these, according to something set in the hash
file, too. Except, there is no corresponding code to load this stuff from
the hashfile (that I can tell, at least).
I've tried adding in my own set, but it still fails to find the right
word. The hash lookup seems to be at fault, now. I'm not sure how it's
keyed, but I have the right letters. Also, the whole KEY_SIZE part seems
massively fubared (and file dependent?!?). My guess is that some
(including the particular hash I'm using) have key sizes of 128. However,
we're running with a ispell using 256. Unfortunately, recompiling to 128
breaks stuff somewhere in that horrid mess. Of course, I'm not even sure
that would fix anything.
If someone else wants to beat on it a while, go ahead. I'd rather start
on an ispell replacement than fight with that code anymore. In fact, if
anyone has spell checking code they think would be a good base, go ahead
and send it to me.
Justin