[KinoSearch] Two things

[Marvin Humphrey]

On Wed, May 06, 2009 at 09:59:48PM -0700, webmasters at ctosonline.org wrote:

> (Except that it is documented as returning a list, rather than an  
> array ref. :-)

:(

Current KS never deals in list context.  When the documentation says "an
array", it means an "array ref".  This was an artifact of me trying to set
things up so that the same documentation text could be shared by multiple
binding languages; "array ref" isn't right in the context of Java, C, etc.  

I might have to do something asinine like use #AN_ARRAY# and swap in the name
of choice for the bindings.  

> I’ve reduced it as much as I can and attached it. (I know, I’ve  
> probably done something stupid and it will be obvious to you.) 

I always start with the assumption that if a smart user has done something
"stupid", there's likely an interface design flaw that shunted them in the
wrong direction -- so I always want to hear about such things.

But in this case, the problem was exactly what you diagnosed: a binding bug.
When called from perl, highlight_spans() was invoking Compiler's no-op default
rather than the subclass method. 

Thanks for the reduced test case, which greatly sped troubleshooting.  The
problem with highlight_spans() has been fixed by r4549; a few more latent
binding bugs following the same pattern have been fixed by r4550; and a couple
memory leaks have been plugged by r4551 -- so I suggest trying r4551.

Marvin Humphrey