An attached thread that starts non-D-fibers can cause the GC to crash while scanning the stack

[rainers]

Recently, I noticed crashes and stalls by Visual D within VS2026, where the GC was scanning huge memory blocks as the stack end address was way above the current stack pointer.

I suspect this happens because VS2026 might use fibers (more than previous versions) that change the stack range without the D runtime being notified about this. If the memory range starting with the current stack pointer up to the stack end address captured at the time of thread attachment contains unmapped memory ranges, an access violation is raised.