Revert "bpo-43693: Add the MAKE_CELL opcode and interleave fast locals offsets. (gh-26396)" by pablogsal · Pull Request #26597 · python/cpython

CC: @ericsnowcurrently @markshannon I have tried to find the error myself and debug this issue for 30 min, but the patch in 631f993 is quite involved and I need to catch up with the rest of changes made to the compiler so I was not able to find it quickly :(

The problem is that PyFrame_LocalsToFast(frame, 1); calls into PyFrame_LocalsToFast where co->co_nlocals has a value of 3, and this causes co->co_cell2arg[i - co->co_nlocals]; to access outside the bounds of the array.

🤖 New build scheduled with the buildbot fleet by @pablogsal for commit 01ff3e0 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

@pablogsal, sorry to impose again! I was watching the buildbots after but ran out of time to wait for them all to finish. Lesson learned.

@pablogsal, sorry to impose again! I was watching the buildbots after but ran out of time to wait for them all to finish. Lesson learned.

No problem! I will try to improve the experience a bit better if I managed to find the time. Will try to propose some improvements so maybe the failures are reported in a comment or something

What's the best way for me to run the address sanitizer locally, to reproduce the failure?

...and thanks for the helpful analysis above.

What's the best way for me to run the address sanitizer locally, to reproduce the failure?

I mentioned it in https://bugs.python.org/msg395321 but I should have mentioned it here as well:

$ export ASAN_OPTIONS=detect_leaks=0:allocator_may_return_null=1:handle_segv=0
$ ./configure --with-address-sanitizer --without-pymalloc
$ make -j -s
$ ./python -m test test_sys_settrace

I used gcc 10 for this, but I think clang works as well as any recent version of gcc.