Daan
Daan
Oh, very interesting! We observed a similar error in the wild with mimalloc v3 .. I blamed it on an out of bounds error in the program but now I...
Ah, I cannot build it at this point. Can you clarify how to repro? I pull'd the latest `dev` and then merged this PR. If I look in the `kklib/mimalloc`...
Ah thanks -- i can try later. Did you upgrade to mimalloc v3.1.3 manually?
No need to upgrade mimalloc. I think i misunderstood that you also upgraded mimalloc :-) I think now it might be a bug in the primitives.
@TimWhiting : maybe try to compile with `-O-1 --debug` and then use `lldb ` to execute. I got an assertion failure on the tag of the reference which is suspicious..
It is important to get this right. To me it is unclear what `getXdgDirectory XdgData kkbuild` returns if no XDG environment variables are set? What is returned if `XDG_DATA_HOME` is...
Thank you for the PR! I was also confused because it is already cast to `uint32_t` so the extra `& 31` should not be needed. Then I remembered a recent...
Ah I see. I cannot check this but if it is indeed the case that the insight dll is injected earlier than mimalloc, then you should try to move the...
Mmm, I'll try to run Nvidia Nsight and try to see what is going on. I guess because nsight runs the application it injects it's dll earlier than anything else...
> It seems like in `_mi_os_alloc_aligned` we initialize `mi_memid_t` to `_mi_memid_none()` which leaves `initially_committed` in uninitialized state? ... I don't think this is it as we reinitialize on success with...