Chong Yidong <cyd@stupidchicken.com> writes:
>> OK, it seems that (vertical-motion 0) is reliable now with your patch,
>> but (vertical-motion 1) will sometimes move point to the beginning of
>> the last word in the line (or so), possibly when the visual line is
>> fairly long.  It looks as if it anticipates the word to be wrapped,
>> even though it isn't.
>
> Please provide a detailed recipe.  (If the recipe refers to
> `posn-point', that would be even better.)

1) Put the following (also attached) text (it should be one logical line)
   into a file:

15:11 *** NAMES snogglethorpe munichlinux Bzek jmaslibre twb xinming fukas melmothX xaiki vinc456 stanis_sh bonii LaoLang_cool akm1 meandtheshell edrx 

2) visit the file with emacs -Q (assuming your patch has been applied),
   and do M-: (setq word-wrap t) to turn on word-wrapping.

3) Now locate the cursor on the first SPC character.  Do M-: (posn-at-point).

   Note that the position is (5 . 0) in characters, and (5*W . 0) in pixels,
   where W is the basic character width (W = 10 for the usual default font).

4) Use `C-f' to move forward one character.  Do M-: (posn-at-point).

   Note that the position is now (7 . 0) in characters, and (7*W . 0) in
   pixels!  This is entirely unexpected, and does not happen if word-wrap is
   nil.

[Because of this funniness, movement with Stefan's line-move-visual patch
also acts weird on this text.]

-Miles