GH-93516: Store offset of first traceable instruction in code object #93769
Conversation
…ute it all the time when tracing.
|
Blocked by #93771 |
|
As an experiment, can you remove the "import opcode" from deepfreeze.py and hardcode the RESUME opcode? We can debate whether that's the right fix, but so far the deepfreeze.py script has been carefully designed not to depend on the details of the target version. |
|
Did you consider putting the new field in the marshaled code object? Then deepfreeze.py wouldn't have to import |
Yes, but I decided against it. |
|
Thanks @markshannon for the PR |
| @@ -20,6 +20,9 @@ | |||
| verbose = False | |||
| identifiers, strings = get_identifiers_and_strings() | |||
|
|
|||
| # This must be kept in sync with opcode.py | |||
| RESUME = 151 | |||
|
|
|||
There was a problem hiding this comment.
Can't this be opcode.opmap["RESUME"]?
There was a problem hiding this comment.
No, that would introduce a dependency on the latest opcode.py which would mean you cannot run deepfreeze.py with e.g. Python 3.10. On Windows and for cross compilations we need to be able to do that. (Windows doesn't have a separate bootstrap Python because it would near-doubel build times.)
We started with this (see #93771).
Avoids having to recompute the offset all the time when tracing.