On 05/20/2010 10:32 PM, Chong Yidong wrote:
> Oh, yes, I forgot.  There is image-cache-eviction-delay as well.  If you
> set this to a sufficiently small number, the test frees the pixmaps as
> required.

Yes, that works, thanks.

> Is there a specific reason you are interested in this problem?  It's
> possible that we need to improve the conditions under which the image
> cache is freed, but it's hard to tell unless we have a use-case in mind.
> (Obviously, the current system assumes that the rate at which images are
> created is much lower than the "usual editing speed".)

Well, I'm toying with a little vector graphics API built on top of 
emacs' svg support. And my first real example is a clock. By leaving it 
running for a few minutes easily makes my X server go out of memory. You 
can find the library (vrend.el) and the example implementing a clock 
(vrend-example-clock.el) attached.

Not sure this clock is a real use-case. But I have a few other ideas in 
mind on how to use it, which all have substantially lower requirements 
on refresh rate, but probably large enough to create lots of pixmaps in 
30 minutes (which seems to be default value for image-cache-eviction-delay)

  anders