"Memory Leak" in $jsilxna.imageMultipliedCache about jsil

[kg]

Yeah, it's probably worth evicting bitmaps from the cache after it reaches a certain size. If you want to take a swing at such an implementation I'll gladly merge it. I'm not sure that a game that fills the cache has any hope of ever performing well, though, so it may not be worth the effort.

Any ways to do this natively? Nope, canvas pretty much sucks. I experimented with doing it via SVG filters, since they can do this natively with hardware acceleration, but every modern browser's SVG implementation is so hopelessly kneecapped and full of bugs that you can't use it for this either. Only option is WebGL, and right now that's not a realistic enough option to use as the standard implementation for the shims. A dual canvas/webgl impl like forplay uses might be cool, though.

Edit: If you provide a test case that hits this bug, that'd also be adequate for me to be able to write a fix against. Right now since none of my code hits it it's unlikely I'd fix it myself.

[kg]

I've updated the image multiply caching system in fd0a850 to reduce the number of cached bitmaps it uses (alpha is no longer handled using the multiply system, it just keys off r/g/b now). This should reduce memory usage somewhat. Still no cache eviction algorithm though.

[kg]

Should be fixed by 2ea2402.