René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"JIT error stubs don't account for peeks","articleBody":"# Crash report\r\n\r\nWhen deciding how many stack items to pop in an `_ERROR_POP_N` stub, we use `_PyUop_num_popped`. However, for opcodes with \"peeked\" items that never get popped, this is incorrect.\r\n\r\nFor example, consider `SET_ADD(2)`. It has three inputs and two outputs. However, both of the bottom inputs are the same as the outputs, and are never really popped. When compiling an error stub for the JIT, using `_PyUop_num_popped` will create `_ERROR_POP_N(3)`, which leaks both references and shrinks the stack too far:\r\n\r\n```py\r\n# 17 Nones to warm up, then a list to raise TypeError:\r\nitems = 17 * [None] + [[]]\r\n{item for item in items}\r\n# Assertion failed: (STACK_LEVEL() \u003e= level), function _PyEval_EvalFrameDefault, file ceval.c, line 966.\r\n```\r\n\r\nThe compiled trace for the comprehension is:\r\n\r\n```\r\n   0 OPTIMIZED: _START_EXECUTOR (0, target=20, operand=0x101644de0)\r\n   1 OPTIMIZED: _MAKE_WARM (0, target=0, operand=0)\r\n   2 OPTIMIZED: _SET_IP (0, target=20, operand=0x1017b85f8)\r\n   3 OPTIMIZED: _CHECK_PERIODIC (0, jump_target=0, operand=0, error_target=13)\r\n   4 OPTIMIZED: _CHECK_VALIDITY (0, jump_target=14, operand=0x1017b85f0)\r\n   5 OPTIMIZED: _ITER_CHECK_LIST (4, jump_target=15, operand=0)\r\n   6 OPTIMIZED: _GUARD_NOT_EXHAUSTED_LIST (4, jump_target=16, operand=0)\r\n   7 OPTIMIZED: _ITER_NEXT_LIST (4, target=16, operand=0)\r\n   8 OPTIMIZED: _STORE_FAST_0 (0, target=18, operand=0)\r\n   9 OPTIMIZED: _LOAD_FAST_0 (0, target=18, operand=0)\r\n  10 OPTIMIZED: _SET_IP (0, target=19, operand=0x1017b85f6)\r\n  11 OPTIMIZED: _SET_ADD (2, jump_target=0, operand=0, error_target=17)\r\n  12 OPTIMIZED: _JUMP_TO_TOP (0, jump_target=1, operand=0)\r\n  13 OPTIMIZED: _ERROR_POP_N (0, target=0, operand=0x14)\r\n  14 OPTIMIZED: _DEOPT (0, target=16, operand=0)\r\n  15 OPTIMIZED: _EXIT_TRACE (0, target=16, operand=0x101644e60)\r\n  16 OPTIMIZED: _EXIT_TRACE (0, target=23, operand=0x101644e70)\r\n  17 OPTIMIZED: _ERROR_POP_N (3, target=0, operand=0x13)\r\n```\r\n\r\n@markshannon, does this make sense to generate new tables for in the cases generator? Not sure what the best option is.\n\n\u003c!-- gh-linked-prs --\u003e\n### Linked PRs\n* gh-126507\n\u003c!-- /gh-linked-prs --\u003e\n","author":{"url":"https://github.com/brandtbucher","@type":"Person","name":"brandtbucher"},"datePublished":"2024-10-31T03:23:55.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":1},"url":"https://github.com/126222/cpython/issues/126222"}