Re: Flyspell loading first time buries buffer and inhibits mode hooks (Win32)

[petermonsson]

On 20 Maj, 12:53, Tony Mc <af...@btinternet.com> wrote:
> Is that literally true? The process isn't easy to imagine.

I mean, yeah, I have to take a shower each time I start Emacs. For
some people that would be a good time to do it anyway, but I start up
Emacs 5 or 6 times a day so it quickly starts getting annoying.

Anyway I've found out why I have the problem, but first I need to
admit that I have forgotten a relevant part of my .emacs file:

(setq default-major-mode 'text-mode) ;; rather than fundamental

Alright, so when I find most files, text-mode will load and flyspell
will load because text-mode is loaded. Apparently this happens all the
time when I start Emacs from explorer. I suppose that *scratch* is one
of the few exceptions that don't depend on text-mode.

In *Messages* I have the following output:
Loading longlines...done
Loading flyspell...done
Starting new Ispell process [default] ...
ispell-init-process: aspell did not output version line

That doesn't tell much. If I comment out (setq default-major-mode
'text-mode) I don't get have problems with ruby-mode, cc-mode, etc.,
but I like text-mode as the default mode much more than fundamental.

Looking at *Messages* after opening a text file I see it has changed
to:
Loading longlines...done
Loading flyspell...done
Starting new Ispell process [default] ...
File mode specification error: (error "aspell did not output version
line")

Hmm, new incompatible version of aspell perhaps? (I check whether the
same problem exists on an Emacs 21.3 on a linux box. There is no
problem and aspell loading is extremely fast compared to my win box)
Aspell version on linux box is 0.50 so there might be a version
incompatibility here. (Note: Aspell 0.50 was released before last
ispell.el release and 0.6 after the last ispell.el release.)

I dive into Ispell.el to see what's wrong. I can't find anything that
has to do with line parsing around that error message, but there are
methods which invoke aspell and wait for it until some timeout limit.
That might be it? My computer is dog slow, but extending the timeout
limit has no effect except for giving the same error much later :(.

What does error do? Stopping whatever the current and parent methods
do. If programming is robust, then the developer frees resources and
returns everything back to the state before the method was run. I
don't think that is what is happening here, although I haven't looked
at the source thoroughly. What happens if I replace the (error ...)
call with nil? Bingo! the annoying behavior dissappears.

Now flyspell won't exactly work like a charm because of that, but I
can live with having to do M-x flyspell-buffer the first time I want
to spell check a buffer and flyspell is now a pleasure and not a pain
to work with.

I don't think other people will have the same error as me, but for the
public service

Best Regards
Peter