Matthieu Dartiailh

Thanks @P403n1x87 I will try to have a look at the recompilation issue.

I tried but cannot reproduce the recompilation issue. Can you ?

So that's really weird. Can you reproduce on 3.11 ?

I tried re-running and I got the same failure on 3.13 but I do not understand why it does not reproduce locally

Thanks for confirming. When I tried for 3.13 I did not use the full harness just tried disassembling the offending classmethod. I guess I will have to do the full...

So I can finally reproduce using: ``` import dis import io from types import CodeType from _pytest import debugging from bytecode import Bytecode, ControlFlowGraph def transform(code): # Round-trip the code...

I have a fix that work locally on the misbehaving case. Let's see how much it helps. Ideally we need a dedicated test to catch such issue faster in the...

So we still have the weird test failure on 3.13 that I do not understand. And 3.11 passed here and failed on your other branch, this one I may ignore...

Thanks for the reproducer I was able to add a minimal test.

This is something I am aware. When rebuilding from the abstract version we put things in order while the Python compiler sometimes produce a different order. So far it did...