Re: Emacs trunk(+AUCTeX) regressions

[Angelo Graziosi]

With a binary search, the changes which cause the problems I see, are 
localized between

CVS of  20090129 15:30  (works fine)

and

CVS of  20090129 15:40  (does not work fine)

i.e.

$ cat src_ChangeLog-20090129_1540-1530.diff
--- /tmp/emacs/src/ChangeLog    2009-02-01 16:24:46.031250000 +0100
+++ emacs/src/ChangeLog 2009-02-01 14:02:47.000000000 +0100
@@ -1,3 +1,26 @@
+2009-01-29  Chong Yidong  <address@hidden>
+
+       * keyboard.c (pending_signals): New var.
+       (poll_for_input, input_available_signal, init_keyboard): Set it.
+       (process_pending_signals): New function.
+
+       * lisp.h (QUIT): Check pending_signals instead of
+       interrupt_input_pending.  Use process_pending_signals.
+
+       * atimer.c (run_timers, alarm_signal_handler): Update
+       pending_signals.
+
+       * process.c (wait_reading_process_output): Use
+       process_pending_signals.
+
+       * sysdep.c (emacs_write): Use process_pending_signals.
+
+       * xterm.c (XTread_socket): Update pending_signals.
+
+       * w32term.c (w32_read_socket): Update pending_signals.
+
+       * w32inevt.c (w32_console_read_socket): Update pending_signals.
+
 2009-01-29  Kenichi Handa  <address@hidden>

        * xftfont.c (xftfont_has_char): New function.


To reproduce the things I have attached a tarball which contains a .tex 
file to be used under Emacs+AUCTeX and a skeleton of my ".emacs" file. 
Indeed something there triggers the problems. As you can see in ".emacs" I 
have (since I was 'children' with Emacs...):

(setq imenu-auto-rescan-maxout 500000)

which seems to cause the problems. If you comment out it, then 
Emacs-builds > 20090129-15:30 seems to work normally. Instead with 
"imenu-auto-rescan-maxout" > 80000 (default 60000, me 500000), when I type 
a key, it is echoized (appears) after: 9 seconds (gulp!) on Cygwin and 1.5 
seconds on GNU/Linux Kubuntu 8.04.02!

As I wrote: it is more than an year I use Emacs23-gtk+AUCTeX, and with 
"(setq imenu-auto-rescan-maxout 500000)", without big problems.

Ralf, I am afraid I can't give a short example to be used with emacs -Q... 
The test case should have many lines and many sectioning (section, 
chapter, subsection...). However the test case I send is simple to use: 
after visiting it with Emacs, one should only type a character a see after 
how many seconds it appear (obviously all this with the skeleton .emacs 
substituting yours temporarly).


Cheers,
Angelo.



On Sun, 1 Feb 2009, Ralf Angeli wrote:
> * Angelo Graziosi (2009-02-01) writes:
>
> > It seems that the changes in Emacs 23.0.60 trunk that have been done
> > between 2009.01.27 and 2009.01.31 cause big regressions on Cygwin and
> > GNU/Linux Kubuntu 8.04.02 (even if here they are weaker), mainly with
> > AUCTeX.
> >
> > Visiting a tex file can take several seconds and when one type a character
> > it takes some time before it is echoized. The effect are greater with
> > bigger file. Also closing a buffer or quitting from Emacs takes some time.
>
> I cannot reproduce this with
> emacs -Q -l auctex circ.tex
> using an anti-aliased DejaVu Sans Mono font or a non-anti-aliased
> Terminal font on Debian sid.  In the anti-aliased case there is a
> noticable lag during scrolling.  But it is not so bad that it shows
> during character insertion.
>
> So please provide a minimal test case starting from emacs -Q (likely
> with a sample LaTeX file) for reproducing the problem.

test_emacs.tar.bz2
Description: BZip2 compressed data