Closed forchid closed 1 month ago
Thank you! I can reproduce the crash and will look at it today.
An interesting finding in the object allocation stress test code.
There were two issues with the code due to the number of objects that were being allocated.
The first issue was allocated objects not being found on the JIT processor stack due to nested function calls. Those roots were not accounted for when that happened, and their in-use memory was deallocated.
The second issue was due to the size of the backing array for the Hash Objeck class. When hashtables spilled, and linked lists became extremely long, the memory manager would blow the stack when marking memory. An interim fix is in place; a longer-term fix is to remove recursion from the memory manager.
Fixed in 2024.10.0
Version info
Version: 2024.9.0 (x86_64 Windows)
Test case
The test source
The test result