https://buildbot.python.org/all/#/builders/1/builds/601

OK (skipped=1)
.
test_threading leaked [9770, 9772, 9768] references, sum=29310
test_threading leaked [3960, 3961, 3959] memory blocks, sum=11880
2 tests failed again:
    test_asyncio test_threading

❯ ./python -m test test_threading -m test_excepthook_thread_None -R :
Run tests sequentially
0:00:00 load avg: 1.38 [1/1] test_threading
beginning 9 repetitions
123456789
.........
test_threading leaked [364, 364, 364, 364] references, sum=1456
test_threading leaked [164, 164, 164, 164] memory blocks, sum=656
test_threading failed

== Tests result: FAILURE ==

1 test failed:
    test_threading

Total duration: 159 ms
Tests result: FAILURE

The problem here is that there is a reference cycle with threading.ExceptHookArgs but structseq objects are not tracked by the GC.

Łukasz, this PR will fix the x86 Gentoo Refleaks 3.x and friends, so I would recommend landing this before the release.

New changeset 7ffcf848df214135abeea7f6c6faa4135fd0928f by Pablo Galindo in branch 'master':
bpo-37126: Allow structseq objects to be tracked by the GC (GH-13729)
https://github.com/python/cpython/commit/7ffcf848df214135abeea7f6c6faa4135fd0928f

You forgot to call PyObject_GC_UnTrack() in structseq_dealloc(), no?

> You forgot to call PyObject_GC_UnTrack() in structseq_dealloc(), no?

Yep, thanks for the catch!

Thanks.

commit 3caf4de6f05f68c3a175f4d8ce870d7a0016622a
Author: Pablo Galindo <Pablogsal@gmail.com>
Date:   Sun Jun 2 21:52:49 2019 +0100

    Call PyObject_GC_UnTrack in structseq dealloc (GH-13751)

--

FYI I also fixed test_threading:

commit cdce0574d03005e27b843fc110c54c99c7a76412
Author: Victor Stinner <vstinner@redhat.com>
Date:   Sun Jun 2 23:08:41 2019 +0200

    bpo-36829: test_threading: Fix a ref cycle (GH-13752)