| 19 Nov 2024 | |||
| patrickb | Coke: I've had a good experience with WinDbg | 22:44 | |
| Don't forget to add --debug to the moar Configure.pl | 22:45 | ||
| 20 Nov 2024 | |||
| [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 | ||
| 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 | ||
| [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 | ||