Repository navigation
Cannot perform frame jump after handled exception #92228
Description
Activity
@dkrystki Do you have a full reproducer?
You haven't said how you are performing the jump.For the presented example I performed jump using PyCharm.
Not sure if I know how to reproduce it using settrace (I can try) but under the hood PyCharm is simply setting f_lineno to perform a jump.Edit: For my library purposes I am setting f_lineno manually and it is causing the same issue
I managed to make a full reproducer:
import sys def fun(): a = 1 # <---- jump location try: b = 1 / 0 except ZeroDivisionError as e: pass c = 3 # current location def trace(frame, event, arg): if event == "line" and frame.f_lineno == 11: frame.f_lineno = 4 return trace sys.settrace(trace) if __name__ == "__main__": fun()Traceback:
Traceback (most recent call last): File "<cwd>/cakeshop.py", line 24, in <module> fun() ^^^^^ File "<cwd>/cakeshop.py", line 11, in fun c = 3 # current location ^ File "<cwd>/cakeshop.py", line 11, in fun c = 3 # current location ^ File "<cwd>/cakeshop.py", line 16, in trace frame.f_lineno = 4 ^^^^^^^^^^^^^^ ValueError: can't jump from within an exception handlerdiff --git a/Lib/test/test_sys_settrace.py b/Lib/test/test_sys_settrace.py index b1c8f6f80a..34126464bc 100644 --- a/Lib/test/test_sys_settrace.py +++ b/Lib/test/test_sys_settrace.py @@ -2217,6 +2217,16 @@ def test_jump_backwards_into_try_except_block(output): raise output.append(6) + @jump_test(7, 1, [1, 3, 6, 1, 3, 6]) + def test_jump_backwards_over_try_except_block(output): + output.append(1) + try: + output.append(3) + 1/0 + except ZeroDivisionError: + output.append(6) + x = 7 + # 'except' with a variable creates an implicit finally block @jump_test(5, 7, [4], (ValueError, 'within')) def test_no_jump_between_except_blocks_2(output):
All works if add anything above or below the assignment.
For what it's worth, this is the problematic result from the reproducer.
Bytecode:
Details
3 0 RESUME 0 4 2 LOAD_CONST 1 (1) 4 STORE_FAST 0 (a) 6 6 NOP 7 8 LOAD_CONST 1 (1) 10 LOAD_CONST 2 (0) 12 BINARY_OP 11 (/) 16 STORE_FAST 1 (b) 11 18 LOAD_CONST 3 (3) 20 STORE_FAST 3 (c) 22 LOAD_CONST 0 (None) 24 RETURN_VALUE >> 26 PUSH_EXC_INFO 8 28 LOAD_GLOBAL 0 (ZeroDivisionError) 40 CHECK_EXC_MATCH 42 POP_JUMP_FORWARD_IF_FALSE 13 (to 70) 44 STORE_FAST 2 (e) 9 46 POP_EXCEPT 48 LOAD_CONST 0 (None) 50 STORE_FAST 2 (e) 52 DELETE_FAST 2 (e) 11 54 LOAD_CONST 3 (3) 56 STORE_FAST 3 (c) 58 LOAD_CONST 0 (None) 60 RETURN_VALUE 62 LOAD_CONST 0 (None) 64 STORE_FAST 2 (e) 66 DELETE_FAST 2 (e) 68 RERAISE 1 8 >> 70 RERAISE 0 >> 72 COPY 3 74 POP_EXCEPT 76 RERAISE 1 ExceptionTable: 8 to 16 -> 26 [0] 26 to 44 -> 72 [1] lasti 62 to 70 -> 72 [1] lastiResult from
mark_stacks():int64_t *stacks = { [0] = 0, [1] = 0, [2] = 3, [3] = 0, [4] = 0, [5] = 3, [6] = 15, [7] = 3, [8] = 3, [9] = 0, [10] = 3, [11] = 0, [12] = 3, [13] = -2, [14] = -2, [15] = -2, [16] = -2, [17] = -2, [18] = -2, [19] = -2, [20] = -2, [21] = -2, [22] = -2, [23] = -2, [24] = -2, [25] = -2, [26] = -2, [27] = -2, [28] = -2, [29] = -2, [30] = -2, [31] = -2, [32] = -2, [33] = -2, [34] = -2, [35] = -2, [36] = -2, [37] = -2, [38] = -2, [39] = -2, };Here,
-2meansUNINITIALIZED. This is a result of the following code inmark_stacks:for (i = 0; i < len; i++) { int64_t next_stack = stacks[i]; if (next_stack == UNINITIALIZED) { continue; } // ... }
After
RETURN_VALUE, the rest of the stacks areUNINITIALIZEDand remainUNINITIALIZEDas they're skipped over. I don't know the correct solution.BTW, bisected to adcd220
Reacted by Damian KrystkiewiczI'm making this a deferred blocker.
I don't think it needs to block the next beta, but I definitely want to fix it before releaseReacted by Damian KrystkiewiczWe need to use the exception handling table to fill in the stacks for exception handlers.
Fiddly, but not too difficult.29 remaining items
The issue of exception handlers is tracked here: #94739
- Repository owner moved this from In Progress to Done in Release and Deferred blockers 🚫
on Jul 11, 2022 The jump from the example I give is from the
pass, which is outside of the except handler.- Repository owner moved this from Done to In Progress in Release and Deferred blockers 🚫
on Jul 11, 2022 I see, will check.
This fails too:
@jump_test(1, 3, [1, 3]) def test_jump_over_return(output): output.append(1) return 12 output.append(3)That should fail.
output.append(3)is unreachableYou mean, the compiler can tell that it's unreachable.
If you allow the debugger to jump around you can do all kinds of impossible things (like get to the pass in your example without an exception being raised). I'm not sure what the semantics of these debugger jumps are, really.
I'm not sure what the semantics of these debugger jumps are, really.
No one is.
In my mind, they are:
- Never crash the interpreter.
- If it is possible to make the jump, then do so.
- Present a meaningful error message if the jump is impossible.
Reacted by Irit KatrielWe understand the issue, and it's the subject of #94739 so I'll close this again (the original problem reported here was resolved).
- Repository owner moved this from In Progress to Done in Release and Deferred blockers 🚫
on Jul 11, 2022
For given code:
Jumping from
current locationtojump locationyieldsValueError: can't jump from within an exception handlerYour environment
Works as expected for python < 3.11