Re: Fwd: Serious performace problems on Windows XP with new(!) GNU Emacs v22 (both patched and unpatched EmacsW32 were tried)

[Peter Tury]

2006/10/19, martin rudalics <address@hidden>:
>  >> > However I see that it is not
>  >> > really modified between the fast (old) and slow (new) Emacses...
>
> Emacs 21.3 had whitespace 3.1 which did not highlight incriminated areas
> and used slightly simpler regexps.  Emacs 22 has whitespace 3.5 which
> uses overlays to highlight such areas.  The overhead for maintaining
> these overlays may grow non-linearly with respect to their number.

Do you mean whitespace-highlight-the-space? I tried it with setting
whitespace-display-spaces-in-color to nil and whitespace-buffer is
still very slow (though a bit faster). It seems for me that
highlighting takes the smaller part (5-10 seconds?) from that 20-25
seconds I measured on my PC. The bigger problem is somewhere else(?).

Especially because "older" v22(!) Emacs (e.g. EmacsW32 of May -- see
my previous letters) was much more faster: they opened the problematic
files in 1-3(?) seconds -- and whitespace.el wasn't really modified
since then!? (I will double check that version again.)

> Your
>
>  > 3. unzip and then visit the attached slowtst.el (I drag&dropped it)
>
> has some 3500 lines which will produce approximately 7000 overlays.

Yes, this is awfully filled up with whitespace errors, since I
prepared it for demonstration. However the problem came for me from
"real life files": some of them took several minutes to open (if their
mode was associated with whitespace):

> Moreover, quitting is inhibited during the scans, hence if
>
>  > 4. M-x whitespace-buffer: it took ~20-25 seconds for me! (I have a
>
> holds - my mileage is about a minute - Emacs will be unresponsive for
> that time.

As far as I know normal font-lock solves this problem by colorizing
only some parts of the file (~what can be seen in the buffer
actually(?)). Would it be too hard to modify whitespace to work in
this way? (Though I still think that the real problem is introduced
outside of whitespace.el).

At least I am happy to see that you could reproduce the error situation.

Thanks,
P