On 5/2/2011 12:44 PM, Jeff Trawick wrote:
>
> That issue affects this type of pattern string:
>
> res = apr_fnmatch("aaa[a*b", "aaa[a*b", 0);
> ABTS_INT_EQUAL(tc, 0, res);
>
> With the old code (1.3.x branch), it does not match. With the new
> code, it does match.
>
> Just sayin', for now. I'll spend more time with the new code.
I did not refer to apr_fnmatch.h. The language describing the various
behaviors may be BSD in origin and require copyright attribution. Or
it may not; symbol names and bit values cannot be copyrighted.
Just sayin - I don't know if the copyright attribution in apr_fnmatch.h
belongs or not right now, but that file isn't properly branded as part
of an ASF project.
The behavior described was wrong; you can reproduce correct behavior
in BSD, Linux, and probably AIX and other shells as well. Do not
refer to any stupidity in apr_fnmatch.h, but refer only to the specs
cited in apr_fnmatch.c. I should have documented all irregularities
and defined all undefined behaviors, but if you find any ambiguity
please let me know!!!
