bug#15045: Point jumps inappropriately around time of Semantic lexing

[Barry OReilly]

>>> It is also very rare.

>> I don't think the bug is rare. I do think that this bug being
>> triggered is rare, since we don't have many non-idle timers in
>> Emacs that trigger redisplay

> That's "rare" in my book.

Three people in this thread independently witnessed the bug. Given the
proportion of Emacs users who follow the mailing lists, calling this
rare is questionable.

> are there others in Emacs core besides the display-time event
> handler?

I have never used display-time-mode, so another timer is involved,
possibly non core. I don't see this bug nearly as often as David's
"every 15 minutes". I would guess 10 times in the past several
months. I'm running Emacs now with the debugging changes Stefan
suggested:

> in Fredisplay, walk the specpdl stack looking for a
> save_excursion_restore where the saved position is different from
> the current value of point in that buffer.

as the change I posted:
http://debbugs.gnu.org/cgi/bugreport.cgi?bug=15045#29

I did not witness the bug yesterday, but I did hit the debug
statements a few times. In those cases the Fprin1 and Fbacktrace calls
printed empty strings, and coincided with Emacs pausing for a few
seconds. I don't know why I'd get empty strings. Maybe there's a flaw
in that debugging code?