On Thu, Nov 13, 2025 at 03:58:50PM -0600, Eric Blake via Libguestfs wrote:
> No need to re-strdup things in a loop of getline - since we are using
> popen to get the output in the first place, we can use the shell to
> hand us just the final line.
>
> Signed-off-by: Eric Blake <[email protected]>
> ---
>
> Even though the POSIX folks ruled today that our "ab"use of a
> getline() loop to determine the final line of du output was
> non-portable [1], I was still miffed at the malloc overhead that our
> workaround for rawhide glibc entailed. So this is the result I came
> up with.
>
> [1] https://www.austingroupbugs.net/bug_view_page.php?bug_id=1953
>
> plugins/linuxdisk/filesystem.c | 28 +++++++---------------------
> 1 file changed, 7 insertions(+), 21 deletions(-)
>
> diff --git a/plugins/linuxdisk/filesystem.c b/plugins/linuxdisk/filesystem.c
> index 283af61a..aa4615e6 100644
> --- a/plugins/linuxdisk/filesystem.c
> +++ b/plugins/linuxdisk/filesystem.c
> @@ -148,7 +148,7 @@ create_filesystem (struct virtual_disk *disk)
> static int64_t
> estimate_size (void)
> {
> - CLEANUP_FREE char *command = NULL, *line = NULL, *lastline = NULL;
> + CLEANUP_FREE char *command = NULL, *line = NULL;
Doh - this line...
> + /* Should only be one line of input. */
> + if (getline (&line, &len, fp) == -1 || ferror (fp)) {
> nbdkit_error ("getline failed: %m");
> pclose (fp);
> + free (line);
...means that this addition is not only unnecessary, but a double-free
bug. Good thing I waited for a review, rather than pushing right away.
> return -1;
--
Eric Blake, Principal Software Engineer
Red Hat, Inc.
Virtualization: qemu.org | libguestfs.org
_______________________________________________
Libguestfs mailing list -- [email protected]
To unsubscribe send an email to [email protected]
