1:46:09Bikethat's more closely analogous to the native unwinder... but also there's no real way to inform it... bleh
1:47:09Bikei guess i could define a new dynenv class specifically for vm frames, that just records the frame pointer to unwind to
1:47:41Bikethen the unwinder can do that without going through longjmp. so we'd still be stack consing dynenvs and all, but it would at least be somewhat cheaper than an unwind protect dynenv
1:57:20Bikeand we wouldnj't be setjmping for every call, so that's nice.