[PATCH] libbb: shrink is_suffixed_with()

[Isaac Dunham]

On Wed, Aug 26, 2015 at 10:18:33AM -0700, Isaac Dunham wrote:
> By my test (Debian Jessie, GCC 4.9.2, glibc shared build), the results are:
> function                                             old     new   delta
> is_suffixed_with                                      61      45     -16
> ------------------------------------------------------------------------------
> (add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-16)             Total: -16 bytes
>    text	   data	    bss	    dec	    hex	filename
>  760460	   2092	   9080	 771632	  bc630	busybox_old
>  760444	   2092	   9080	 771616	  bc620	busybox_unstripped
> ---
>  libbb/compare_string_array.c | 15 +++++----------
>  1 file changed, 5 insertions(+), 10 deletions(-)
> 
> diff --git a/libbb/compare_string_array.c b/libbb/compare_string_array.c
> @@ -35,17 +35,12 @@ char* FAST_FUNC is_prefixed_with(const char *string, const char *key)
>   */
>  char* FAST_FUNC is_suffixed_with(const char *string, const char *key)
>  {
> +	char *ret;
>  
> +	ret = strstr(string, key);
> +	if (ret && strcmp(ret, key))
> +		ret = NULL;
> +	return ret;
>  }

I forgot to check/update the documentation; the behavior is slightly
different in the case of an empty key.
As a side effect of the implementation, it returns NULL if key is "",
unless string is also "".
It's probably not going to be smaller if this change is not kept.

Regardless which behavior is used, is_suffixed_with() can be used to check
that config files in a directory are named *.conf or for similar purposes;
I think standard modprobe has this behavior, so  for example the file
/etc/modprobe.d/file.conf.ignore
will be ignored. If there's anywhere that we check this, I can't see it.

To prevent the function from being optimized out, I wrote a small applet
which I'm attaching.

Thanks,
Isaac Dunham
-------------- next part --------------
A non-text attachment was scrubbed...
Name: suftest.c
Type: text/x-csrc
Size: 838 bytes
Desc: not available
URL: <http://lists.busybox.net/pipermail/busybox/attachments/20150826/0d838a57/attachment.c>