github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
Kaiepi for IO::Socket::INET connections, what should i do on EAGAIN when SO_SNDTIMEO/SO_RCVTIMEO is used with the socket? 01:32
Xliff m: class A { }; my A $a = Nil; 03:33
evalable6
timotimo yesterday i found out that out of 466k collectables in this one heap snapshot file, about 261k collectables have only a single outgoing reference 11:42
today, i want to find out if there's potential for deduplication at all
28.5k collectables had only two outgoing refs, probably not worth as much, though i should give it a try anyway 11:48
Kaiepi uhhhhh 21:07
moarvm may potentially be vulnerable to ROP
openbsd has a feature called retguard that protects against it, but can be disabled by passing -fno-return-protector to clang and that fixes nearly all the page traps i get when i try to run perl6 21:08
s/page trap/trace trap/ 21:10
timotimo does it work fine when the jit is turned off? 21:13
it probably has to write its own return protection data somewhere for it to not explode every time it returns from inside jit code 21:14
Kaiepi yes, this is with the jit turned off and with -fno-ret-protector enabled 21:16
er 21:17
the jit's enabled
timotimo and with ret protector turned on, does turning off the jit make things work again? 21:18
Kaiepi yes
timotimo OK
the return protector stuff is a CPU feature that has the same API on all OSes? 21:19
or do we have to build platform-specific support for it?
Kaiepi -fno-ret-protector is openbsd specific 21:30
so probably platform specific depending on what security features they have 21:31