Okay, I ran bash under gdb and got a backtrace which perhaps points to
readline. While it's hard to lay blame with a malloc error, this fits
with my experience as bash would often crash immediately after exiting
a command and returning to the prompt.

Josh: Do you have a specialized ~/.inputrc? I do and I wonder if that
could be what's triggering the bug for us. A lesser possibility could
be a customized prompt. Have you changed your PS1 from Debian's
default?

My PS1='\h:\W\$ '

My .inputrc follows
==============

# Time-stamp: <2013-09-17 23:27:36 ben>                 -*- conf -*-
# This file is read by Bash when the user first logs in
# or if the user runs "bind -f ~/.inputrc" (or hits C-x C-r)

# Ben prefers tcsh style history completion for M-p, M-n.
"\en": history-search-forward
"\ep": history-search-backward

# Keep cursor column when using C-p, C-n
# (Ben finds this VERY useful: exploit referential locality)
set history-preserve-point on

# Don't leave lines in the history edited after hitting return.
# (Too many times I've been unable to find commands that I Control-u'd.)
set revert-all-at-newline on

# Readline should not prevent me from binding ^W.
set bind-tty-special-chars off

# ^w wipes a region instead of the standard unix word rubout.
# Now I can use ^w to cut and ^y to paste, just like in emacs.
"\C-w": kill-region

# M-w copies the region
"\ew": copy-region-as-kill

# ^u should clear everything, not just delete to the left!
"\C-u": kill-whole-line

# I like to see the file types during tab-completion (like ls -F).
set visible-stats on

# Tab completion should put a slash at the end of a symlink to a directory
set mark-symlinked-directories on

# Tab completion should not show dot files unless I type a dot first.
set match-hidden-files off


#### Experiment: September 17, 2013 ####
# Should normal C-p (or up arrow) also do history-search-backwards?
# Let's see if I like it.
"\C-n": history-search-forward
"\C-p": history-search-backward
"\e[B": history-search-forward
"\e[A": history-search-backward
"\eOB": history-search-forward
"\eOA": history-search-backward





#### Experiment: September 17, 2013 ####
# I'm not sure that this is useful, but it certainly doesn't seem harmful.
# I'll try it out and see if it bugs me.
#
# When performing completion in the middle of a word, do not insert
# characters from the completion that match characters after point in
# the word being completed, so portions of the word following the
# cursor are not duplicated.
set skip-completed-text on


# #### Experiment: November 10, 2009 #####
# #### Disabled on Ursula because Bash 3.2 doesn't have shell-*-word. 4/2010
# # M-f, M-b, M-d, M-DEL refer to *shell* words rather than English words.
# # For example, in:  mv a\ b\ c\ d z\ y\ x\ w
# # "a b c d" would be one word and "z y x w" would be another word.

# "\ef": shell-forward-word
# "\eb": shell-backward-word
# "\ed": shell-kill-word
# "\e\C-?": shell-backward-kill-word
# "\e\C-h": shell-backward-kill-word

# # Also, bind the original English functions to shifted M-F and M-B.
# # Note that we can't access the shifted kill-word keys because
# # M-D is a VT cursor escape sequence and DEL can't be shifted.
# "\eF": forward-word
# "\eB": backward-word



################### CRUFT TO REMEMBER #####################
# Instead of ringing the bell, show all completion possibilities? NO! BAD!
# Ben was testing. I've left this in here to remind myself that I don't like it.
#set show-all-if-ambiguous on

# 8-bit keystrokes should be interpreted as meta, not latin1.
# (You can to disable this for the next input character with ^V.)
# DISABLED APRIL 28th 2006; Ben is moving to UTF-8 now...
# INSTEAD USE xterm's metaSendsEscape resource.
#set convert-meta on

# Mark modified history lines with an asterisk. (UGLY but USEFUL.)
#
# REASONING:
# Sometimes when I'm searching in my history, the command I'm looking
# for isn't there anymore. The reason is that I had started editing
# the line previously and then used C-n or C-p to go to a different
# line. Unlike tcsh, bash does *not* reset the line to be unedited
# automatically if you don't execute the edited line. By marking
# modified lines, at least I'll know the line was edited and can hit
# M-r to reset it manually.
#
# mark-modified-lines on


-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org

Reply via email to