github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
nwc10 good *, #moarvm 07:08
jnthn morning o/ 10:23
tellable6 2021-02-07T09:43:59Z #moarvm <nwc10> jnthn paste.scsys.co.uk/594542 -- new-disp t/02-rakudo/15-gh_1202.t ASAN excitement
nwc10 \o 10:24
also, I'd not seen ASAN since then, but there are two slightly different traces whereby it can do a NULL pointer deref in src/core/interp.c:4361 10:25
MasterDuke wow, thought i'd give gcc's static analyzer a go at the moarvm build. 32k log file written so far (but it does create very large ascii diagrams) and it seems to have hung on src/strings/shiftjis.o (well, 100% cpu and res is currently 15gb and rising) 10:47
been going for 15min now 10:48
jnthn Maybe the use switch/case in shift_jis_cp_to_index :) 10:54
MasterDuke i killed it. unfortunately doing so kills the whole make, so i need to see if there's a way to skip certain files 10:57
bunch of `warning: dereference of NULL ...` and `warning: double-‘free’ of ...` 11:02
i can follow some of these dereference warnings, but i don't understand the double-free ones 11:29
here's one about MVM_spesh_pea_destroy_deopt_info gist.github.com/MasterDuke17/64794...49c584a508 if anybody wants to take a look 11:32
what would be the right thing to do here github.com/MoarVM/MoarVM/blob/mast...ray.c#L836 if s_repr_data is NULL here github.com/MoarVM/MoarVM/blob/mast...#L813-L814 ? 11:40
nine MasterDuke: that's on gcc 10 I assume? 12:55
MasterDuke yeah
10.2.0 12:56
nine waiting for 11 should be worth it :0
The author claims a much lower false positive rate and loads of improvements in general compared to the alpha version in gcc 10 12:58
MasterDuke yeah, and i see you can annotate functions as malloc/free pairs, which i assume would be good for MVM_malloc/MVM_free/etc 12:59
looks like gcc 11 should be released soon, guess i'll wait until then to revisit the analyzer 13:02
though my question about VMArray's copy_elems still stands
MasterDuke rebased github.com/MoarVM/MoarVM/pull/1426 and broke apart the commits into some smaller chunks 16:10
nine MasterDuke: is the VMArray question purely hyptothetical, or do you see that irl? 16:34
MasterDuke hypothetical 16:41
nine nqp: my @a := 1, 2, 3; my %b := nqp::hash; %b<a> := 1; nqp::splice(@a, %b, 1, 0) 16:54
nqp-m: my @a := 1, 2, 3; my %b := nqp::hash; %b<a> := 1; nqp::splice(@a, %b, 1, 0)
n: my @a := 1, 2, 3; my %b := nqp::hash; %b<a> := 1; nqp::splice(@a, %b, 1, 0)
Darn...camelia is down. Anyway, it's not hypothetical anymore. This ^^^ actually segfaults due to the NULL pointer dereference. 16:57
MasterDuke: great spot!
MasterDuke credit to the gcc analyzer 17:03
lizmat and yet another Rakudo Weekly News hits the Net: rakudoweekly.blog/2021/02/08/2021-...-remaster/ 17:11

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!