[patch] unfreed memory in cli_command_loop

[patch] unfreed memory in cli_command_loop

[msnyder]

[-- Attachment #1: Type: text/plain, Size: 139 bytes --]

Minor memory leak.  I checked to make sure that readline makes
and keeps its own copy of this string, so it's OK for us to
free our copy.


[-- Attachment #2: event-top.txt --]
[-- Type: text/plain, Size: 2069 bytes --]

2007-06-28  Michael Snyder  <msnyder@access-company.com>

	* event-top.c (cli_command_loop): Prompt string can (and should)
	be freed after call to readline (Coverity).  Also move local var
	declarations into block where they are used.

Index: event-top.c
===================================================================
RCS file: /cvs/src/src/gdb/event-top.c,v
retrieving revision 1.50
diff -p -r1.50 event-top.c
*** event-top.c	9 Feb 2007 23:45:35 -0000	1.50
--- event-top.c	29 Jun 2007 02:00:03 -0000
*************** rl_callback_read_char_wrapper (gdb_clien
*** 186,204 ****
  void
  cli_command_loop (void)
  {
-   int length;
-   char *a_prompt;
-   char *gdb_prompt = get_prompt ();
- 
    /* If we are using readline, set things up and display the first
       prompt, otherwise just print the prompt. */
    if (async_command_editing_p)
      {
        /* Tell readline what the prompt to display is and what function it
           will need to call after a whole line is read. This also displays
           the first prompt. */
!       length = strlen (PREFIX (0)) + strlen (gdb_prompt) + strlen (SUFFIX (0)) + 1;
        a_prompt = (char *) xmalloc (length);
        strcpy (a_prompt, PREFIX (0));
        strcat (a_prompt, gdb_prompt);
        strcat (a_prompt, SUFFIX (0));
--- 186,206 ----
  void
  cli_command_loop (void)
  {
    /* If we are using readline, set things up and display the first
       prompt, otherwise just print the prompt. */
    if (async_command_editing_p)
      {
+       int length;
+       char *a_prompt;
+       char *gdb_prompt = get_prompt ();
+ 
        /* Tell readline what the prompt to display is and what function it
           will need to call after a whole line is read. This also displays
           the first prompt. */
!       length = strlen (PREFIX (0)) 
! 	+ strlen (gdb_prompt) + strlen (SUFFIX (0)) + 1;
        a_prompt = (char *) xmalloc (length);
+       make_cleanup (xfree, a_prompt);
        strcpy (a_prompt, PREFIX (0));
        strcat (a_prompt, gdb_prompt);
        strcat (a_prompt, SUFFIX (0));

Re: [patch] unfreed memory in cli_command_loop

[Daniel Jacobowitz]

On Thu, Jun 28, 2007 at 07:05:32PM -0700, Michael Snyder wrote:
> Minor memory leak.  I checked to make sure that readline makes
> and keeps its own copy of this string, so it's OK for us to
> free our copy.

Seems fine, but I'm not really sure when that cleanup is going to be
called.  We're outside the command loop here so it hasn't recorded a
watermark to clean back to.  Does it ever get run?  If not, you could
use alloca and avoid the whole issue.

-- 
Daniel Jacobowitz
CodeSourcery

Re: [patch] unfreed memory in cli_command_loop

[Michael Snyder]

> On Thu, Jun 28, 2007 at 07:05:32PM -0700, Michael Snyder wrote:
> > Minor memory leak.  I checked to make sure that readline makes
> > and keeps its own copy of this string, so it's OK for us to
> > free our copy.
>
> Seems fine, but I'm not really sure when that cleanup is going to be
> called.  We're outside the command loop here so it hasn't recorded a
> watermark to clean back to.  Does it ever get run?  If not, you could
> use alloca and avoid the whole issue.

Whoa, you're right, thanks for the catch.

Re: [patch] unfreed memory in cli_command_loop

[msnyder]

> On Thu, Jun 28, 2007 at 07:05:32PM -0700, Michael Snyder wrote:
>> Minor memory leak.  I checked to make sure that readline makes
>> and keeps its own copy of this string, so it's OK for us to
>> free our copy.
>
> Seems fine, but I'm not really sure when that cleanup is going to be
> called.  We're outside the command loop here so it hasn't recorded a
> watermark to clean back to.  Does it ever get run?  If not, you could
> use alloca and avoid the whole issue.

Withdrawn, will resubmit.