From: <me...@ho...> - 2005-06-08 18:26:26
|
The signal fixes brought another problem to the surface: gencgc alloc floods the system with interrupts when SUB-GC allocates or after a gc triggered in a WITHOUT-GCING form. I think cheney gc does not have this problem as it clears the gc trigger before invoking the gc. Surprisingly only the non-threaded builds failed when exceeding the maximum interrupt nesting depth. Threaded builds did not trip over simply because the with-mutex form in sub-gc protected them from consing. Gabor |