My fuzzer created a situation where gc-heap-committed-unused is temporarily larger than gc-heap-allocated. This caused an assertion failure:
###!!! ASSERTION: Invalid value for js-main-runtime-gc-heap-committed-unused-ratio: 'false', file aboutMemory.js, line 0
7,711,104 B ── js-main-runtime-gc-heap-allocated
1,083,008 B ── js-main-runtime-gc-heap-arena-unused
2,097,152 B ── js-main-runtime-gc-heap-chunk-clean-unused
4,837,376 B ── js-main-runtime-gc-heap-chunk-dirty-unused
15,728,640 B ── js-main-runtime-gc-heap-committed
8,017,536 B ── js-main-runtime-gc-heap-committed-unused
103.97% ── js-main-runtime-gc-heap-committed-unused-ratio [?!]
0 B ── js-main-runtime-gc-heap-decommitted
Presumably a regression from http://hg.mozilla.org/mozilla-central/rev/f5a301fe9ba5. Bug 748440 comment 1 even anticipated the issue. Why was this reporter changed from a percentage to a ratio?
> Why was this reporter changed from a percentage to a ratio?
For the purposes of telemetry, I thought the ratio was a more interesting measurement.
We should just get rid of this assertion. I looked for it and couldn't find it, but I clearly didn't look hard enough.
Created attachment 620710 [details] [diff] [review]
However turned test_aboutmemory.xul permaorange on OS X, so backed out:
And landed again with a test fix. https://hg.mozilla.org/integration/mozilla-inbound/rev/0528b55a2616