github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
MasterDuke it's interesting, on my remove_candidates branch heaptrack shows 20m more calls to allocation functions and 4.7m more temporary allocations, but peak heap and rss are almost identical 09:26
so just a lot more churn, but not more memory being used in total 09:27
MasterDuke oh, valgrind (and heaptrack) show a leak on my branch, haven't figured out what's leaking yet. gist.github.com/MasterDuke17/42853...171b92ff5b 11:41
MasterDuke ok, removing isn't the problem, it leaks even if no candidates are ever removed 13:07
well, i added what i believe was a missing free of `cands_and_arg_guards->spesh_candidates` in MVMStaticFrameSpesh.c's gc_free, but still the same leak 13:19
brrt good * 14:43
nwc10 good *, brrt 15:05
it is sunny outside. I'm going back outside
(just popping inside to collect some inspiration from the beer fridge) 15:06
it is sunny "this week"
MasterDuke it is the remove_candidates branch that introduces the leak, it's not there on master 15:26