This failure happens about every 2-3 runs on the MIPS buildbot. We see:
stress/new-largeish-contiguous-array-with-size.js.ram-size-10000000: Exception: Error: heap too big after forced GC: 1368418
Locally, I can reproduce and get a different value:
# i=1; while jsc --forceRAMSize=10000000 JSTests/stress/new-largeish-contiguous-array-with-size.js ; do echo $i; i=$((i+1)); done
Exception: Error: heap too big after forced GC: 1355035
It is unclear to me whether this is a sign of a bug or that the test needs to check against a higher maximum.
It looks like on different runs, I get one of the following progression of heap sizes:
- 3359463 before GC and 158823 after GC
- 3359495 before GC and 158855 after GC
- 2159159 before GC and 558839 after GC
- 1759047 before GC and 958887 after GC
- 2559143 before GC and 1358903 after GC (failing the test)
For some of them, I get slight variations with the last 2 digits being different.