On Mon, Jan 09, 2006 at 01:53:13PM +0100, Agustin Martin wrote:
> c) Run again flyspell-region on the same region
>    => if: Buffer french-LIZEZMOI has no process

Just noticed that this becomes evident when ispell patch for
"flyspell needs ispell-change-dictionary twice" is applied. When using
current plain ispell.el from CVS this error message does not appear, but
things are also buggy; although silently, dict is not changed.

> 
> > However, if calling flyspell-accept-buffer-local-defs twice does no
> > harm, it would be fine to move that call into flyspell-region.
> 
> However, you are right, (flyspell-word) is the key here and if instead of
> (c) you run flyspell-word over a word the same error appears. I think we can
> handle this from ispell.el as part of the
> ispell-change-dictionary+double-call problem making sure
> flyspell-last-buffer is nil'ed on dict change if flyspell is loaded. this
> way first call to flyspell-word would re-run
> flyspell-accept-buffer-local-defs and skip that for further calls from the
> same buffer. No need to add an explicit call in flyspell-small-region or
> flyspell-region.
> 
> This might also fix the original problem Piet described, but in a different
> way (that looks compatible with the current fix, no need for changes there).

I am attaching a possible patch for that. It is done on top of the (not yet
installed) "flyspell needs ispell-change-dictionary twice" patch. It is the
incremental one,

[ispell.el.ispell-internal-change-dictionary+incremental.diff]

(ispell-internal-change-dictionary)
 - Clear out flyspell-last-buffer if flyspell is loaded and dict is changed,
   so (flyspell-accept-buffer-local-defs) is re-run for the new values.

I am also attaching the full patch for current ispell.el from CVS, so it can
be tested better if previous patch is not yet installed when reading this
mail.

[ispell.el.ispell-internal-change-dictionary+full.diff]

-- 
Agustin