github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
02:53 linkable6 left, linkable6 joined 03:24 leont left 07:11 frost-lab joined 07:52 domidumont joined 09:01 Altai-man joined 09:09 sena_kun joined
dogbert11 interesting, it's actually possible, with some tweaks, to have t/spec/S17-promise/lock-async-stress2.t generate a SEGV 09:09
09:10 Altai-man left
dogbert11 here: github.com/MoarVM/MoarVM/blob/mast...rp.c#L2109 09:13
(gdb) bt 09:15
#0 MVM_interp_run (tc=0x0, tc@entry=0x557b34ecfd50, initial_invoke=0x557b33f0dde0, initial_invoke@entry=0x7f43913a19a0 <thread_initial_invoke>, invoke_data=0x557b33f0dde0,
invoke_data@entry=0x7f43913a19a0 <thread_initial_invoke>, outer_runloop=outer_runloop@entry=0x0) at src/core/interp.c:2109
#1 0x00007f43913a1a7d in start_thread (data=0x557b34ecfc20) at src/core/threads.c:91 09:16
#2 0x00007f4390e04609 in start_thread (arg=<optimized out>) at pthread_create.c:477
09:24 zakharyas joined
nine So obj is NULL? 09:25
dogbert11 yes 09:35
note that I'm on HEAD
in order to get the SEGV I had to set #define MVM_NURSERY_SIZE 21000 09:37
nine: more importantly, have you received your new gear? 09:39
10:14 MasterDuke left
nine Not yet. Should be getting it this week though 10:19
Can you catch it in rr and re-trace where that NULL is coming from? 11:16
dogbert11 no, unfortunately it turns out that rr doesn't work under VirtualBox 11:27
11:42 patrickb joined 11:53 frost-lab left 12:47 leont joined 13:04 patrickb left 13:08 Altai-man joined 13:11 sena_kun left 13:32 patrickb joined 13:34 Kaeipi left, Kaeipi joined 14:50 zakharyas left
Geth MoarVM/inlining_of_curcode_op: 490ee80d20 | (Stefan Seifert)++ | 3 files
Make curcode OP inlinable

curcode was marked :noinline as tc->cur_frame->code_ref would point at the inliner, not at the inlinee, thus giving the wrong result. As jnthn++ noticed though, we have the right code_ref easily available in a register, since we need it for several ops already. So just turn the curcode into a set, reading from the register holding the inlinee's code_ref.
15:07
MoarVM: niner++ created pull request #1405:
Make curcode OP inlinable
15:08
nine Wow...this is one of the very, very rare "write the patch, compile and pass every test you throw at it on the very first try" occasions. I even went back and broke it intentionally to see if its even active and yes it is. 15:10
[Coke] nine++ 15:25
lizmat nine++ 15:55
leont Can you still buy a lotery ticket today? :-p 15:59
Geth MoarVM: 490ee80d20 | (Stefan Seifert)++ | 3 files
Make curcode OP inlinable

curcode was marked :noinline as tc->cur_frame->code_ref would point at the inliner, not at the inlinee, thus giving the wrong result. As jnthn++ noticed though, we have the right code_ref easily available in a register, since we need it for several ops already. So just turn the curcode into a set, reading from the register holding the inlinee's code_ref.
16:37
MoarVM: 98fccebf98 | niner++ (committed using GitHub Web editor) | 3 files
Merge pull request #1405 from MoarVM/inlining_of_curcode_op

Make curcode OP inlinable
16:55 patrickb left
lizmat reason not to bump MoarVM and NQP now ? 17:07
17:09 sena_kun joined 17:11 Altai-man left, zakharyas joined 17:35 domidumont left
tobs .tell MasterDuke I have no experience with Windows, sadly. 18:49
tellable6 tobs, I'll pass your message to MasterDuke
18:52 zakharyas left 19:23 travis-ci joined
travis-ci MoarVM build failed. niner 'Merge pull request #1405 from MoarVM/inlining_of_curcode_op 19:23
travis-ci.org/MoarVM/MoarVM/builds/750052840 github.com/MoarVM/MoarVM/compare/8...fccebf983e
19:23 travis-ci left 19:59 MasterDuke joined
MasterDuke . 20:19
tellable6 2020-12-16T18:49:26Z #moarvm <tobs> MasterDuke I have no experience with Windows, sadly.
20:39 leont left 20:58 zakharyas joined 21:03 sena_kun left 21:37 leont joined 21:57 zakharyas left 22:12 MasterDuke left 23:16 MasterDuke joined
MasterDuke ugh. `D:\a\1\MoarVM\src\core/threadcontext.h(5): fatal error C1083: Cannot open include file: 'gmp.h': No such file or directory`. i haven't figured out how to fix this 23:17
jnthn I'm guessing it correctly checked it out under 3rdparty/gmp? 23:29
MasterDuke i assume, the linux/mac builds succeeded 23:31
i tried explicitly adding `3rdparty/gmp/gmp.h` to github.com/MoarVM/MoarVM/blob/mast...le.in#L257 but that dies also with something like `make: don't know how to make target '3rdparty/gmp/gmp.h'` (on all the CI targets, enough it's fine locally) 23:33
dev.azure.com/MoarVM/MoarVM/_build...4c3b9def93
jnthn ah, you're doing it on the CI not locally, so it's harder to poke 23:37
MasterDuke yep
jnthn Figured I'd try a Windows build (didn't for ages), and... 23:39
Updating submodules .................................... List form of pipe open
not implemented at tools/update-submodules.pl line 71.
MasterDuke istr some chat about that a little while ago? 23:40
jnthn /I3rdparty/gmp is in the includes list, at lesat
(going by the azure output)
hah, apparently it's a Perl bug fixed long ago
Hm, I have 5.16 :)
sih 23:48
*sigh
c:\consulting\moarvm\src\core/str_hash_table_funcs.h(55) : error C2275: 'size_t'
: illegal use of this type as an expression
src\main.c : see declaration of 'size_t'
that probably means that I've got an ancient MSVC++ too, which we no longer build on
And that I'm not figuring out tongiht 23:49
MasterDuke looks like azure uses visual studio 2019 23:50
appveyor uses 2017
it gives the same `Cannot open include file: 'gmp.h': No such file or directory` error 23:51
but yeah, time to sleep here...
jnthn Yeah, same here. 23:52