[pacman-dev] [PATCH] Fix several memleaks, mostly related to errors handling.

[Dan McGee]

On Nov 26, 2007 5:07 PM, Xavier <shiningxc at gmail.com> wrote:
> On Mon, Nov 26, 2007 at 03:49:14PM -0600, Aaron Griffin wrote:
> > On Nov 24, 2007 5:02 AM, Xavier <shiningxc at gmail.com> wrote:
> > > That's the first thing I thought about, but I don't know if that is possible.
> > > There are two lists that can only be freed after the error happens : *data
> > > and deps (asked could be freed before).
> > > Or maybe it is possible to do something like :
> > > MALLOC(miss, sizeof(pmdepmissing_t), FREELIST(*data);  FREELIST(deps);
> > >   RET_ERR(PM_ERR_MEMORY, -1));
> > >
> > > Though, there is also the "goto cleanup;" that would be skipped, and so the
> > > "alpm_list_free(list);", so that's yet another call to add there.
> > > So that would duplicate the "FREELIST(deps)'" and "alpm_list_free(list);"
> > > calls.
> >
> > While this one has already been merged, you could still use the macro
> > by simply throwing something like (void)0 in the third param. Or even
> > pm_errno = BLAHBLAH.
> > This way we still get the output from the macro, even if we don't get
> > the real failure+return.
> >
>
> I see, but how do we detect if it failed or not? Using 0, it doesn't seem to
> be possible. Maybe by setting pm_errno as you said, or ret.
>
> So maybe something like this :
>
> MALLOC(miss, sizeof(pmdepmissing_t), ret = -1);
> if(ret == -1) {
>         /* free the lists */
>         pm_errno = blabla;
>         goto cleanup;
> }
>
> ?
>
> I don't know how MACRO handle arguments, so I wasn't sure this was possible.

I think it looks OK. Remember that a macro gets handled during
preprocessing and isn't a function call at all, so you can do a few
things you wouldn't normally be able to. Want to check? Look into
using "gcc -E <sourcefile>", which just spits the preprocessed code
back at you so you can check things.

-Dan