Re: [PATCH] sort.c: Fix minor memory leak, "files" is never free'd

[Joey Degges]

On Tue, Feb 16, 2010 at 11:23 AM, Jim Meyering <j...@meyering.net> wrote:

> Joey Degges wrote:
> > On Tue, Feb 16, 2010 at 5:05 AM, Eric Blake <e...@byu.net> wrote:
> >
> >> According to Joey Degges on 2/16/2010 2:02 AM:
> >> > Hello,
> >> >
> >> > At sort.c:3271 'files' is allocated but it is not free'd before main
> >> exits:
> >> >     files = xnmalloc (argc, sizeof *files);
> >>
> >> Thanks for the patches.  However, calling free() immediately before
> exit()
> >> is a lint-like activity - it is actually SLOWER to explicitly free
> memory
> >> rather than just exiting and letting the OS cleanup reclaim the memory
> as
> >> part of process death.  If we accept patches like this, it will be to
> make
> >> other leak detections easier, but as such, it should probably be
> properly
> >> guarded by #if LINT or something similar to make it apparent that it is
> >> only needed when looking for leaks and not in the common case.
> >
> >
> > Thanks for your insight -- I was not aware of 'lint' before. I have
> > reformatted the patch with #ifdef lint so that this will only be used if
> > gcc-warnings is enabled. If this looks good I will also resubmit the
> other
> > two patches.
> >
> >>From 0018a314269bc8a9b89e82be2cbf17a08d28f297 Mon Sep 17 00:00:00 2001
> > From: Joey Degges <jdeg...@gmail.com>
> > Date: Mon, 15 Feb 2010 23:30:31 -0800
> > Subject: [PATCH 1/3] sort.c: Fix minor memory leak, 'files' is never
> free'd
>
> Thanks for caring, but at least in my book, this is not even a "minor
> leak".
> A more apt one-line summary would be "sort: free memory allocated in main"
>
> >  src/sort.c |    5 +++++
> >  1 files changed, 5 insertions(+), 0 deletions(-)
> >
> > diff --git a/src/sort.c b/src/sort.c
> > index 481fdb8..a2cba05 100644
> > --- a/src/sort.c
> > +++ b/src/sort.c
> > @@ -3692,6 +3692,11 @@ main (int argc, char **argv)
> >    else
> >      sort (files, nfiles, outfile);
> >
> > +#ifdef lint
> > +  if (nfiles != 0)
> > +    free (files);
> > +#endif
>
> Also, the above will malfunction when nfiles is initially 0,
> since we set file = &minus and "nfiles = 1".
> With your proposed addition, that would make sort
> call free(&minus), which would probably segfault.
>
> Considering that freeing "files" is not at all important,
> I think it would be better to tell whatever tool you're
> using that this particular case is not a problem.
> In valgrind, you can add a so-called "suppression".
>

Thank you for the comments. I know that it would easier to just suppress the
errors and move on but I would like to see this fixed.

This fixes the issues where nfiles is 0 and also when --files0-from is
specified. In addition, this may speed up the case where nfiles is 0 since
files would never be free'd.

 diff --git a/src/sort.c b/src/sort.c
index 481fdb8..435f6be 100644
--- a/src/sort.c
+++ b/src/sort.c
@@ -3166,6 +3166,7 @@ main (int argc, char **argv)
   bool posixly_correct = (getenv ("POSIXLY_CORRECT") != NULL);
   bool obsolete_usage = (posix2_version () < 200112);
   char **files;
+  char **files_ptr;
   char *files_from = NULL;
   struct Tokens tok;
   char const *outfile = NULL;
@@ -3268,7 +3269,7 @@ main (int argc, char **argv)
   gkey.month = gkey.reverse = false;
   gkey.skipsblanks = gkey.skipeblanks = false;

-  files = xnmalloc (argc, sizeof *files);
+  files_ptr = files = xnmalloc (argc, sizeof *files);

   for (;;)
     {
@@ -3559,6 +3560,7 @@ main (int argc, char **argv)
         {
           size_t i;
           free (files);
+          files_ptr = NULL;
           files = tok.tok;
           nfiles = tok.n_tok;
           for (i = 0; i < nfiles; i++)
@@ -3651,7 +3653,6 @@ main (int argc, char **argv)
     {
       static char *minus = (char *) "-";
       nfiles = 1;
-      free (files);
       files = &minus;
     }

@@ -3692,6 +3693,12 @@ main (int argc, char **argv)
   else
     sort (files, nfiles, outfile);

+#ifdef lint
+  free (files_ptr);
+  if (files_ptr == NULL)
+    readtokens0_free (&tok);
+#endif
+
   if (have_read_stdin && fclose (stdin) == EOF)
     die (_("close failed"), "-");