With the upvar2 patch the ARM box is reporting different stats from what are in the file. The actual ARM stats accord with my pre-upvar2 tree when I select only that one test, but they don't agree with my tree if I don't select out the single test. (I have not tested post-upvar2.) Andreas suggests this is a problem with the oracle, which seems plausible based on my quite meager knowledge of it.
Should we patch trace-test.js to fix the orange? /be
Interesting; I have seen similar problems with other tests, though they've often gone away after some tweaking.
I changed the stats to pass on tinderbox and overall, but we should still figure out why an individual run fails.
Obsolete with the removal of tracejit.