|
08:19
vrurg left
08:20
vrurg joined
13:03
vrurg_ joined
13:05
vrurg left
|
|||
| patrickb | It seems that for a thread gc status the value `MVMGCStatus_NONE | MVMSuspendState_SUSPEND_REQUEST` is invalid (looking at github.com/MoarVM/MoarVM/blob/main...e.c#L337). | 17:28 | |
| Still state is set in the debugserver here github.com/MoarVM/MoarVM/blob/main...ver.c#L697 | 17:29 | ||
| I'm not entirely clear on the invariants of the GC status and how the different status' and suspendstates are meant to interact. | 17:30 | ||
| Can someone (timo?) clarify? | |||
| (I'm AFK soon, but will backlog!) | |||
| timo | hm, could be that one is just tricky to hit in normal operation as another code path that checks the gc status for having to enter gc from interrupt would be running more often or something? seems like it should be allowed in the first link you posted | 17:55 | |
|
18:44
vrurg_ left
|
|||
| patrickb | Ah, I think I got it. MVM_GC_enter_from_interrupt is able to deal with MVMSuspendState_SUSPEND_REQUEST. So I'll just need to adapt the if condition. Nice! Thanks for the quick help! | 18:55 | |
| wait that's wrong. I hit that code path with a plain SUSPEND_REQUEST. I'll probably need to cas it to UNABLE | SUSPEND_REQUEST. Does that sound about right? | 19:12 | ||
|
20:29
vrurg joined
|
|||