[ESS-bugs] ess-beginning-of-function - bug in {R and S}-mode

[Martin Maechler]

I've just renamed <ess>/etc/indent-wrong.R
to etc/R-ESS-bugs.R
and added an example which demonstrates the bug:

This bug applies to both R- and S-mode , i.e. typically when
using *.R and *.S (or *.q) source files.

C-M-a and C-M-e   and all other ess-function-* commands fail 
for things like

  myfun <-
    ## bla1
    ## bla2
    ## bla3
    ## bla4
    function(x) {
     .........
     .........
    }

as long as your (point) is inside the  '## bla.' comment lines

John Chambers has been using this syntax extensively,
and actually S4 has defined that the '## bla?" lines are made
into a ``dummy help page'' as long as no real help page exists.
and IIRC, this even works in S-plus {but not in R}.

The error message happens from this part 
in (defun ess-beginning-of-function) :

    (while (not done)
      (if (re-search-backward ess-function-pattern (point-min) t)
	  nil
	(goto-char init-point)
	(error "Point is not in a function.")

	))

where the culprit is really ``hidden in'' the variable
ess-function-pattern which is initialized from e.g.,
ess-R-function-pattern from ess-cust.el

But I don't see how, since line 644 {ess-cust.el} has

   "\\(\\(\\s-\\|\n\\)*\\s<.*\\s>\\)*" ; whitespace, comment

i.e. *is* supposedly looking for comment lines before the 
     'function(..)' keyword.

Does anyone have an idea?

Martin