gh-128089: Bump the PYC magic number

[JelleZijlstra]

• Issue: NotImplementedError raised by annotationlib (?) when importing classes derived from NamedTuple #128089

[vstinner]

LGTM

[JelleZijlstra]

Unrelated failure in test_glob

======================================================================
FAIL: test_selflink (test.test_glob.SymlinkLoopGlobTests.test_selflink)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/runner/work/cpython/cpython-ro-srcdir/Lib/test/test_glob.py", line 543, in test_selflink
    self.assertIn(path, results)
    ~~~~~~~~~~~~~^^^^^^^^^^^^^^^
AssertionError: 'dir/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/file' not found in {'dir/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/file', 'dir/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/file', 'dir/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/link/file'}

[vstinner]

I re-ran the failed job.

[JelleZijlstra]

Seems like another PR bumping the magic number just went in. Is it still worth it to merge this change @hroncok? I could bump the number further but then the patch wouldn't apply cleanly to 3.14a3 anyway.

[Eclips4]

LGTM, but it seems to have a merging conflict now. By looking at d2f1d91, it appears that d2f1d91 should use Python 3.14a4 instead of Python 3.14a3, is that right?

[hroncok]

Ah. We cannot backport a version bump that would be equal to a4, or else we'll have the same problem when a4 is out.

[hroncok]

But we can take this, make it 3610. The main branch already had 3611 it seems.

[JelleZijlstra]

Main is still on 3610, so it's possible the next alpha will get released with that number.

[vstinner]

Since the magic number was bumped to 3610 by another change, I suggest to not bump the magic number again. Maybe just change the magic number comment to mention annotationlib?

[vstinner]

Or if you still want to increase the magic number, maybe the NOT_TAKEN instruction should get 3611, and annotationlib should get 3610. I don't know :-(

[hroncok]

Main is still on 3610...

cpython/Include/internal/pycore_magic_number.h

Line 279 in 78ffba4

#define PYC_MAGIC_NUMBER 3611

[JelleZijlstra]

Oh right, the last comment is for 3610 but the actual number is 3611.

[JelleZijlstra]

Apparently 3610 was accidentally skipped (#128115). Which gives us the opportunity to retcon 3610 into being the annotations change.

[befeleme]

Hello, is there any chance this could be merged? Or what's the plan forward?

[vstinner]

The NEWS entry is outdated:

Bump the PYC magic number for an internal change in annotate functions. The
magic number should have been increased in release 3.14a3. Users running
3.14a3 with PYC files generated using earlier alphas may run into problems.

I suggest to remove it.

[vstinner]

I enabled auto-merge.

[vstinner]

@befeleme: magic number 3610 is now assigned to annotationlib VALUE_WITH_FAKE_GLOBALS. You can backport this change to alpha3 if you want.

[vstinner]

PR merged, thanks @JelleZijlstra.

[befeleme]

@befeleme: magic number 3610 is now assigned to annotationlib VALUE_WITH_FAKE_GLOBALS. You can backport this change to alpha3 if you want.

Thank you