Mirror of the gdb-patches mailing list
 help / color / mirror / Atom feed
From: Andrew STUBBS <andrew.stubbs@st.com>
To: GDB Patches <gdb-patches@sourceware.org>
Subject: command_line_input() not re-entrant
Date: Thu, 30 Mar 2006 14:29:00 -0000	[thread overview]
Message-ID: <442BD6F1.8070804@st.com> (raw)

Hi,

I have discovered a problem in the GDB command line reading code.

command_line_input() uses a static buffer to hold the current command. 
This means that it is not properly re-entrant - commands that contain 
other commands, such as user defined commands, are not handled safely.

In practice the only real trouble I have observed is with user defined 
commands that use $arg0 etc. because these parameters are never copied 
out of the original string, so are overwritten the next time 
command_line_input() is invoked. Even then, this is not normally a 
problem because command_line_input() is not normally needed within a 
user-defined command - it has already been read. It is only a problem 
when the user defined command contains a source command.

The problem may be reproduced as follows:

Create three files:

a1
---8<---------->8-----
source a2
abcdef qwerty
---8<---------->8-----

a2
---8<---------->8-----
define abcdef
   echo 1: <<<$arg0>>>\n
   source a3
   echo 2: <<<$arg0>>>\n
end
---8<---------->8-----

a3
---8<---------->8-----
#################################################################
---8<---------->8-----

Then run the following command:

$ gdb -nx -q -x a1 -batch
1: <<<qwerty>>>
2: <<<######>>>

Both 1: and 2: should have been the same. As you can see the contents of 
a3 have overwritten the value of $arg0 in abcdef. For some reason I 
haven't discovered (and probably boils down to dumb luck) I can't 
reproduce the problem when entering a1 interactively - I have to source it.

I am happy to write the patch to fix this but I am wondering how. There 
seem to be two possible ways:

1. Make command_line_input() re-entrant. Perhaps drop the static buffer 
and malloc a new string each time. Free it through a clean-up.

2. Have setup_user_args() copy the data and adjust the clean up to free 
the copied data.

Any preferences or other suggestions?

Andrew Stubbs


             reply	other threads:[~2006-03-30 13:05 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-03-30 14:29 Andrew STUBBS [this message]
2006-04-04 10:27 ` [PATCH] allow nested sourced commands Andrew STUBBS
2006-04-04 10:34   ` Andrew STUBBS
2006-04-04 19:58     ` Michael Snyder
2006-04-05 10:05       ` Andrew STUBBS
2006-04-05 18:48         ` Michael Snyder
2006-04-06  9:49           ` Andrew STUBBS
2006-04-06 13:40             ` Daniel Jacobowitz
2006-04-07 11:17               ` Andrew STUBBS
2006-04-07 13:18                 ` Daniel Jacobowitz
2006-04-07 13:33                   ` Andrew STUBBS

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=442BD6F1.8070804@st.com \
    --to=andrew.stubbs@st.com \
    --cc=gdb-patches@sourceware.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox