00:11 gfldex left 00:22 lizmat joined 00:32 gfldex joined 00:40 gfldex left 00:42 gfldex joined
[Coke] coleman: I think we're skipping the release this month. 01:09
Not running blin at the moment.
and away from my win machine.
(debug, windbg - Thanks!)
f 01:31
09:02 sena_kun joined 09:34 sena_kun left 09:56 sena_kun joined 10:24 sena_kun left
[Coke] patrickb: I've installed windbg - I was able to get it to attach to the failing moarvm process... and then it seems to hang and not die (have tried continuing with go/step, nothing) - see either a single or bunch of threads, but no exception 17:30
timo were you able to see if it also starts a process of its own perhaps? 20:13
Geth MoarVM/debugserver_loaded_file_notifications: 79676c282b | (Timo Paulssen)++ | 3 files
"really loaded" flag for breakpoint file entries
 20:19
[Coke] Sorry, yes. Was able to go back and stuff all the command line args in and kick it off. got something, maybe. 20:25
timo cool 20:26
[Coke] gist.github.com/coke/d1daf4bb96892...208149bfae
timo huh. that's what it would look like if we haven't "turned unwinding off" 20:28
[Coke] added the callsite info for 0x7
timo should be able to MVM_dump_backtrace(0x2d346060200) and print ((MVMObject *)0x2d34d2e7028)->st[0] to get a bit of info about the involved object 20:31
able to call MVM_dump_blabla etc etc 20:32
not that i know how to do that exactly in windbg
[Coke] WARNING: Continuing a non-continuable exception 20:37
ntdll!RtlRaiseStatus+0x3d:
00007ffa`276ffa4d 65488b0c2560000000 mov rcx,qword ptr gs:[60h] gs:00000000`00000060=????????????????
timo that's an odd spelling for that thing 20:38
[Coke] The print is complaining "Couldn't resolve error at..."
timo so i think we should be able to break before the longjmp happens with a condition that checks if we're in jitted code, which i think tc->jit_return_address being nonzero would tell us 20:42
does the windbg you are using have the "go back" button? i.e. time-travelling debugger support? 20:44
[Coke] it's advertising time traveling. 20:46
I can also set a breakpoint somewhere and rerun it
time travel options greyed out
timo you probably have to start it with time traveling turned on before it can be used
"start it" being the target program 20:48
[Coke] I'm going to have to play with this a bit, I'm afraid. 20:51
timo i guess at the point we're at inside the unwind attempt we just can't run shit because the stack is borken? 20:52
patrickb Just to be extra sure, does the Moar source contain this commit: github.com/MoarVM/MoarVM/commit/38...13712e3806 20:53
to enable time travelling you need to tick some checkbox in the run program dialog (before starting the debugging) 20:54
[Coke] no on the commit (rebuilding), understood on the time travel 20:57
timo oh, well, that commit is really necessary haha 20:58
[Coke] typo in that commit: "it's *second*" should be "its *second*" 21:01
hoping for a completely different error on this rebuild. :) 21:02
same error in rakudo build process. 21:11
time travel complaining about someone else using a text file..
timo ah
you still had some process running
"text" there is like "text section" and refers to a binary
yes, that is very funny 21:12
21:21 sena_kun joined 22:01 sena_kun left
[Coke] looks like no change with that commit 22:33
still dying going through moar!MVM_repr_bind_attr_inso
timo that explosion comes from trying to bind an attribute on an object that doesn't have the capability to hold attributes 22:34
i think we do that in some place with a "try" just to see if it would work?
[Coke] launching with time travel, I get a popup that shows the recording command - options are "stop and debug" or "cancel" 22:35
I assume S&D is correct?
afk 22:36

Powered by Cro and the Raku® Programming Language.

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

Thank you!