bug#56682: locked narrowing

[Stefan Monnier]

> And another thought: did you try using format-mode-line as the means
> to get the line number from which you could start?

Yes, but in my tests, it wasn't any faster.

BTW, while I have now found which code flushes the base_line cache
during redisplay (there are a few different places, in
`redisplay_window` and `try_scrolling` (not sure why it needs to be
done at more than one place, tho)),
I still haven't found where that cache gets flushed or ignored for
the case where it's used for `format-mode-line`.  That code doesn't
seem to pay attention to BEG_UNCHANGED nor even to the changes
in narrowing.

Oh... wait... aha!  I still don't know how it handles buffer
modifications, but I now know that it just fails to pay attention to
changes to the narrowing: so, it's no surprised that I couldn't find the
corresponding flushing code, it's just missing:

    emacs -Q lisp/subr.el
    M->
    M-: (list (format-mode-line "%l")
              (save-restriction (narrow-to-region (+ (point-min) 4000) 
(point-max))
                                (format-mode-line "%l")))

returns twice the same line number.


        Stefan