#12 python-mode kills arrow in gdb (gud.el)

[Matthias Klose]

[this is http://python.org/sf/606250, reopened]
forwarded from http://bugs.debian.org/159628 ]

Once python-mode.el has been loaded, all GDB buffers
stop showing the little arrow that lives in the
"fringe" on the left side of an emacs21 frame. To be
specific, the arrow appears and then disappears right
away, each time a 'step' or 'next' command is run. I
traced the problem down to the unconditional hooking of
'comint-output-filter-functions, python-mode.el line
3472 to be exact:

(add-hook 'comint-output-filter-functions
'py-pdbtrack-track-stack-file)

I haven't tested it fully, but I've found that if I
take this line out, GDB can show its arrow once more.
There is code in python-mode.el to insert this hook in
a buffer-local fashion only in the '*Python*' buffer..
maybe that should be reviewed and used instead of this
global hook?

(took forever to figure out it was python-mode causing
the problem..)

Followup:

This one is still broken for me, in python-mode 4.40. (I'm
the submitter of the original debian bug report).

Here are some instructions to reproduce the problem:

===============
574:warner@cinla% cat hello.c
#include <stdio.h>

int main(void)
{
printf("hello\n");
printf("world\n");
printf("\n");
return 0;
}
575:warner@cinla% gcc -g -o hello hello.c
576:warner@cinla% ./hello
hello
world

577:warner@cinla% cat foo.py
#! /usr/bin/python

print "hiya"
578:warner@cinla% emacs -q

M-x gdb hello RET # starts GUD session
(gdb) b main RET
(gdb) run RET # GDB stops on "hello".
Note
arrow in fringe.
C-x C-f foo.py RET # loads python-mode.el
C-x b *gud-hello* RET # switch back to GDB buffer
(gdb) next RET # GDB stops on "world".
Note
lack of arrow.
(gdb) next RET # arrow flickers and
disappears each time
===================

emacs-version reports "GNU Emacs 21.3.1 (i386-pc-linux-gnu,
X toolkit) of 2003-10-31 on raven, modified by Debian"
this is package emacs21, version 21.3+1-4 (from unstable)
python-elisp is version 2.3.2.91-1, with py-version at
$Revision: 4.40 $

let me know if there is anything I can do to further track
it down.
thanks!
-Brian

[David Kĺgedal]

Logged In: YES
user_id=1260741

I usually work around this by applying this patch:

--- python-mode.el~ 2005-04-08 16:45:30.000000000 +0200
+++ python-mode.el 2005-06-21 11:53:55.000000000 +0200
@@ -1334,7 +1334,7 @@
(setq overlay-arrow-string "=>")
(set-marker overlay-arrow-position (py-point 'bol)
(current-buffer))
(setq py-pdbtrack-is-tracking-p t))
- (overlay-arrow-position
+ (nil ;; overlay-arrow-position
(setq overlay-arrow-position nil)
(setq py-pdbtrack-is-tracking-p nil))
))

The problem is that the comint filter function is called
even when running gud, and sinc it doesn't detect a python
line to show, it disables the overlay arrow.

[Gustav Hållberg]

Logged In: YES
user_id=1310184

I did some testing around this about two years ago, and I
then concluded that the right fix is similar to what
d_kagedal suggests, but that you want the deactivation to be
run iff py-pdbtrack-is-tracking-p:

--- python-mode.el~ 2004-10-21 14:27:30.000000000 -0700
+++ python-mode.el 2005-07-09 23:03:43.417759376 -0700
@@ -1334,7 +1334,7 @@
(setq overlay-arrow-string "=>")
(set-marker overlay-arrow-position (py-point 'bol)
(current-buffer))
(setq py-pdbtrack-is-tracking-p t))
- (overlay-arrow-position
+ (py-pdbtrack-is-tracking-p
(setq overlay-arrow-position nil)
(setq py-pdbtrack-is-tracking-p nil))
))

[Christian Schlotter]

Logged In: YES
user_id=1440711

Is the patch added to the last comment the official
workaround? I would like to submit a package to my favorite
distribution which fixes this problem, as I find it highly
annoying.

Regards
Christian