Re: Excessive refontification when setting jit-lock-context-unfontify-pos

[martin rudalics <rudalics@gmx.at>]

 > in AUCTeX I am extending the region to be fontified by
 > `font-lock-default-fontify-region' backwards in order to cater for
 > large multiline constructs.  In order to force refontification with
 > jit-lock, `jit-lock-context-unfontify-pos' is set to the start of such
 > a multiline construct.  Now when this is done it seems that the
 > affected regions are refontified over and over again.

That's the purpose of setting `jit-lock-context-unfontify-pos' when
`jit-lock-contextually' is non-nil.  jit-lock will trigger refontifying
the _entire visible_ text below `jit-lock-context-unfontify-pos' after
`jit-lock-context-time' seconds.

 >
 > Here is a testcase:
 >
 > (progn
 >   (defun my-font-lock-fontify-region (beg end &optional loudly)
 >     (setq jit-lock-context-unfontify-pos (- beg 1000))
 >     (message (format "Fontifying ... (buffer: %s, beg: %d)"
 > 		     (prin1-to-string (current-buffer)) beg))
 >     (font-lock-default-fontify-region beg end loudly))
 >   (find-library "lisp-mode")
 >   (setq font-lock-fontify-region-function 'my-font-lock-fontify-region))
 >
 > After executing the form lisp-mode.el should be opened and the
 > `font-lock-fontify-region-function' will be set to the new one defined
 > in the code which always sets `jit-lock-context-unfontify-pos'.  If
 > you now scroll to the bottom of the buffer with `C-v' the function
 > obviously gets called a lot.  But once the bottom of the buffer is
 > reached I'd expect that to stop unless something changed in the
 > buffer.  However, only some activity like scrolling or activating the
 > minibuffer will make the fontification function be called again.

Everything's correct.  Scrolling will move you into some text marked as
not fontified and will trigger fontification from there which will, due
to your setting, eventually mark the text as not fontified again, which
will ...

 > This does not happen in Emacs 21, so it looks like this is a
 > regression.  I got a bit tangled up in the font-lock and jit-lock code
 > and hope that somebody has an idea of what might be going wrong.

It's the desired behavior.  `jit-lock-context-unfontify-pos' should be
set by jit-lock only.  You are supposed to set this only in extreme
cases and exercise care to never set this repeatedly.  Why can't you use
font-lock's new extend-region stuff for this purpose?