STEPS TO REPRODUCE:
1) Load http://material.ly/products/547
2) Hold down the right arrow key for a bit. Or just hit that key every so often.
ACTUAL RESULTS: As the page loads images, memory usage continues to climb, getting up over a gigabyte for me pretty easily. Opening about:memory and telling it to minimize memory usage releases all the memory.
EXPECTED RESULTS: GC or CC should happen sometime...
Also, if there is any way to work around this in the page, the page author would love to know in http://stackoverflow.com/questions/9366825/firefox-specific-memory-increases-in-heavy-ajax-app
Olli jiggered with how XHR stuff interacts with the CC in the last week, but I assume some random person on the internet probably isn't running Nightly.
Justin, does this sound like an image-suck bug? The scenario here sounds vaguely like bug 679775.
I can look into how GC/CC scheduling is working (or otherwise) on this page if it isn't just badness about the way images are being handled.
Kyle might also have some idea.
Sounds an awful lot like Bug 681479 to me.
> ACTUAL RESULTS: As the page loads images, memory usage continues to climb, getting up
> over a gigabyte for me pretty easily.
bz, were you seeing uncompressed image memory usage, or just compressed? If you ask about:memory just to GC/CC (not minimize memory usage), does it release all compressed in addition to uncompressed?
(Minimize memory usage should release all background uncompressed images unconditionally.)
I tried reproducing myself, but my internet connection here is way too slow.
(In reply to Andrew McCreight [:mccr8] from comment #2)
> Olli jiggered with how XHR stuff interacts with the CC in the last week, but
> I assume some random person on the internet probably isn't running Nightly.
That affects only certainly alive XHRs.
And yes, this sounds like Bug 681479.
> bz, were you seeing uncompressed image memory usage, or just compressed?
gc/cc releases some of the uncompressed memory, but not all.
And now I can't seem to get this to go above 600MB or so... :(
If some but not all the uncompressed memory is released on gc/cc, then this sounds like a combination of bug 681479 and bug 679775.