Re: master 5141234acf: Refactor the webp cache code to allow usage by gif_load, too

[Lars Ingebrigtsen <larsi@gnus.org>]

Po Lu <luangruo@yahoo.com> writes:

> The object might be garbage collected, and another one could be consed
> with the same type tag and address as the original.

It seems rather unlikely -- especially since it's also checking that the
index is the expected one.

> So we should probably mark that data upon garbage collection.

For the imagemagick cache, we're just using a cookie, that's basically
what we're doing here, too.  Perhaps sxhashing the spec and using that
instead would be fine, but it's slower, and we don't really care that
much about correctness here -- we're just using this to speed things up,
after all.

So I dunno.  My only worry was that not gc-protecting it might lead to a
segfault somewhere.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no