xfree/make_cleanup memory leak pattern?

xfree/make_cleanup memory leak pattern?

[Khoo Yit Phang]

Hi,

I noticed the following code pattern in various places in GDB:
	script = compute_python_string (cmd->body_list[0]);
	...
	xfree (script);

and also others like:
	full_path = tilde_expand (filename);
	...
	cleanup = make_cleanup (xfree, full_path); 

If I understand correctly, wouldn't these patterns lead to a memory leak if the user manages to trigger a SIGINT (with immediate_quit in effect) at the "..." in between the memory allocation and xfree/make_cleanup? It's a tiny window in most cases, but something like compute_python_string could take some time.

Yit
January 10, 2012

Re: xfree/make_cleanup memory leak pattern?

[Jan Kratochvil]

On Wed, 11 Jan 2012 01:25:06 +0100, Khoo Yit Phang wrote:
> I noticed the following code pattern in various places in GDB:
> 	script = compute_python_string (cmd->body_list[0]);
> 	...
> 	xfree (script);
> 
> and also others like:
> 	full_path = tilde_expand (filename);
> 	...
> 	cleanup = make_cleanup (xfree, full_path); 

immediate_quit will cause:
	fatal ("Quit");

which calls throw_exception in the end so all the intermediate cleanup
functions get called.  You can trace/debug it yourself.

I do not see a leak there.  Could you provide some debugging output proving
the memory is being leaked?


Thanks,
Jan

Re: xfree/make_cleanup memory leak pattern?

[Jan Kratochvil]

On Wed, 11 Jan 2012 01:30:35 +0100, Jan Kratochvil wrote:
> I do not see a leak there.  Could you provide some debugging output proving
> the memory is being leaked?

Sorry, this was bogus reply.

Yes, I agree.  immediate_quit is not well handled.


Thanks,
Jan

Re: xfree/make_cleanup memory leak pattern?

[Tom Tromey]

>>>>> "Jan" == Jan Kratochvil <jan.kratochvil@redhat.com> writes:

Jan> Yes, I agree.  immediate_quit is not well handled.

I think it is only used in selected spots.

I wonder if we could enhance the exception checker to look for problems
in this area.

Tom