github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
Geth_ MoarVM: ddlws++ created pull request #1271:
Fix utf8-c8 memory usage
02:24
discoD that's my pr if you've got questions/comments/hate. 02:31
nwc10 good *, #moarvm 11:20
nine Darn....I'm pretty sure that sc_idx is not correct: $4 = {sc_forward_u = {forwarder = 0xe3bc340, sc = {sc_idx = 238797632, idx = 0}, st = 0xe3bc340}, owner = 1, flags = 16, size = 48} 14:47
100 % reproducible, always in a call stack involving MVM_profile_heap_take_snapshot 14:48
jnthn Is 0xe3bc340 valid? 14:50
(as a memeory address)
nine (gdb) p *c 14:51
$7 = {sc_forward_u = {forwarder = 0xb1e2e80, sc = {sc_idx = 186527360, idx = 0}, st = 0xb1e2e80}, owner = 1, flags = 16, size = 48}
(gdb) p *c->sc_forward_u->forwarder
$8 = {sc_forward_u = {forwarder = 0xb1e2eb0, sc = {sc_idx = 186527408, idx = 0}, st = 0xb1e2eb0}, owner = 1, flags = 16, size = 48}
0xb1e2e80 doesn't look like an STable either 14:52
MasterDuke our ui64toa and i64toa are taken from github.com/miloyip/itoa-benchmark, but assuming our names match, there are faster versions 14:57
nwc10 nine: ASAN and/or valgrind don't say much/ 14:59
?
nine nwc10: well valgrind does report the invalid read, but at that point it's already too late 15:06
dogbert17 hmm, something is wrong with the heap profiler - MoarVM panic: Adding item to worklist already freed in gen2 21:55
dogbert17 it looks like this: gist.github.com/dogbert17/76bde0a8...3f441d9f62 23:00