On 5/2/2011 5:18 PM, Jeff Trawick wrote:
>
> Index: strings/apr_fnmatch.c
> ===================================================================
> --- strings/apr_fnmatch.c (revision 1098590)
> +++ strings/apr_fnmatch.c (working copy)
> @@ -152,7 +152,7 @@
> }
> else if (**pattern == '?') {
> /* Optimize '?' match before unescaping **pattern */
> - if (!**string || (!slash || (**string != '/')))
> + if (!**string || (slash && (**string == '/')))
> return APR_FNM_NOMATCH;
> result = 0;
> goto fnmatch_ch_success;
>
> fail if end of string being tested
> fail if we're trying to match a '/' but this is a pathname (wildcard
> '?' can't match '/' in that case)
Yup - that was where I went astray, I had moved typical '?'
processing into fnmatch_ch and constructed the test incorrectly
for its new context. Glad we arrived at the same fix!
I fixed the one nested assignment alert, all that remains is the
'conditional expression is constant' out of the "for (; 1;)" loop
with no apparent fix.
