gh-149481: skip FOR_ITER inline specialization for Python __next__

[NekoAsakura]

Thanks for the careful bisect and the empirical fix from @savannahostrowski. ❤️

Benchmark

Benchmark	unpatched	patched	Δ
for_iter_dict_items	185.8 ms ± 1.6 ms	187.0 ms ± 2.8 ms	+0.68%
for_iter_dict_keys	154.1 ms ± 2.1 ms	153.9 ms ± 2.6 ms	-0.11%
for_iter_dict_values	148.2 ms ± 1.0 ms	146.9 ms ± 2.8 ms	-0.85%
for_iter_set	179.1 ms ± 1.9 ms	179.0 ms ± 2.2 ms	-0.07%
for_iter_reversed	136.6 ms ± 1.2 ms	136.0 ms ± 1.9 ms	-0.47%
for_iter_enumerate	201.1 ms ± 2.9 ms	201.1 ms ± 2.4 ms	-0.04%
for_iter_zip	230.6 ms ± 1.2 ms	227.9 ms ± 1.3 ms	-1.15%
for_iter_list	114.2 ms ± 1.8 ms	115.6 ms ± 2.2 ms	+1.26%
for_iter_tuple	104.4 ms ± 1.5 ms	104.7 ms ± 1.7 ms	+0.35%
for_iter_range	137.6 ms ± 2.5 ms	137.8 ms ± 1.9 ms	+0.13%
Geometric mean			-0.03%

Root cause analyse (from opus 4.7)

When __next__ is Python-defined, tp_iternext is the generic slot_tp_iternext slot wrapper that re-enters the eval loop via vectorcall_method. Calling it through the new _ITER_NEXT_INLINE (which captures tp_iternext as a baked-in operand and lives in a tight _JUMP_TO_TOP loop) is correct functionally, but each new outer-iteration class causes _GUARD_TYPE_ITER to fail and warm a side-exit. After ~SIDE_EXIT_INITIAL_VALUE (4000) hits at each side-exit, a side trace forms there; _EXIT_TRACE jumps trace-to-trace via TIER2_TO_TIER2, which doesn't re-check the eval-breaker. With MAX_CHAIN_DEPTH=4 traces possible, the chain accumulates ~4×4000+4002 ≈ 20k inner hits before the system spirals into a tier-2-only loop with no signal checkpoints — matching the observed exact threshold. The pre-PR _FOR_ITER_TIER_TWO path goes through _PyForIter_VirtualIteratorNext, which uses an indirect dispatch on Py_TYPE(iter_o)->tp_iternext and does not produce a per-type guard, so no side-trace storm forms. Falling back to that path for slot iterators sidesteps the issue while keeping _ITER_NEXT_INLINE for C-level iterators (dict, set, enumerate, zip, reversed, etc.), which the existing test_for_iter_direct_* tests exercise.

• Issue: JIT hang in xml.etree.iterparse #149481

[savannahostrowski]

Yep, this was pretty much the exact fix I had locally. Thanks for the quick turnaround on this @NekoAsakura ❤️

I'll wait to see if anyone else wants to have a look.

[markshannon]

Looks good. Thanks for doing this.

[StanFromIreland]

Updating to fix some errors we introduced on the main branch.

[miss-islington-app]

Thanks @NekoAsakura for the PR, and @savannahostrowski for merging it 🌮🎉.. I'm working now to backport this PR to: 3.15.
🐍🍒⛏🤖 I'm not a witch! I'm not a witch!

[bedevere-app]

GH-149523 is a backport of this pull request to the 3.15 branch.

[bedevere-bot]

⚠️⚠️⚠️ Buildbot failure ⚠️⚠️⚠️

Hi! The buildbot aarch64 Android 3.15 (tier-3) has failed when building commit 5cb915d.

What do you need to do:

1. Don't panic.
2. Check the buildbot page in the devguide if you don't know what the buildbots are or how they work.
3. Go to the page of the buildbot that failed (https://buildbot.python.org/#/builders/2113/builds/2) and take a look at the build logs.
4. Check if the failure is related to this commit (5cb915d) or if it is a false positive.
5. If the failure is related to this commit, please, reflect that on the issue and make a new Pull Request with a fix.

You can take a look at the buildbot page here:

https://buildbot.python.org/#/builders/2113/builds/2

Failed tests:

• test_urllib2
• test_urllibnet

Failed subtests:

• test_getcode - test.test_urllibnet.urlopenNetworkTests.test_getcode
• test_ftp_error - test.test_urllib2.HandlerTests.test_ftp_error

Summary of the results of the build (if available):

==

Click to see traceback logs

Traceback (most recent call last):
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/test/test_urllibnet.py", line 107, in test_getcodeself.assertEqual(e.exception.code, 404)
^^^^^^^^^^^^^^^^AttributeError: 'URLError' object has no attribute 'code'
----------------------------------------------------------------------
Ran 12 tests in 0.018s
FAILED (errors=1, skipped=10)
Traceback (most recent call last):
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 1531, in ftp_open
host = socket.gethostbyname(host)
socket.gaierror: [Errno 7] No address associated with hostname
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/test/test_urllib2.py", line 815, in test_ftp_error
urlopen("ftp://www.pythontest.net/")
~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 489, in open
response =self._open(req, data)
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 506, in _open
result =self._call_chain(self.handle_open, protocol, protocol +'_open', req)
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 466, in _call_chain
result = func(*args)
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 1533, in ftp_openraise URLError(msg)
urllib.error.URLError: <urlopen error [Errno 7] No address associated with hostname>
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/test/test_urllib2.py", line 817, in test_ftp_errorself.assertEqual(raised.reason,
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^f"ftp error: {exception.args[0]}")
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^AssertionError: gaierror(7, 'No address associated with hostname') != 'ftp error: 500 OOPS: cannot change directory:/nonexistent'
----------------------------------------------------------------------
Ran 81 tests in 0.043s
FAILED (failures=1, skipped=3)
Traceback (most recent call last):
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/test/test_urllibnet.py", line 107, in test_getcodeself.assertEqual(e.exception.code, 404)
^^^^^^^^^^^^^^^^AttributeError: 'URLError' object has no attribute 'code'
----------------------------------------------------------------------
Ran 12 tests in 0.024s
FAILED (errors=1, skipped=10)
Traceback (most recent call last):
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 1531, in ftp_open
host = socket.gethostbyname(host)
socket.gaierror: [Errno 7] No address associated with hostname
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/test/test_urllib2.py", line 815, in test_ftp_error
urlopen("ftp://www.pythontest.net/")
~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 489, in open
response =self._open(req, data)
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 506, in _open
result =self._call_chain(self.handle_open, protocol, protocol +'_open', req)
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 466, in _call_chain
result = func(*args)
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/urllib/request.py", line 1533, in ftp_openraise URLError(msg)
urllib.error.URLError: <urlopen error [Errno 7] No address associated with hostname>
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/data/user/0/org.python.testbed/files/python/lib/python3.15/test/test_urllib2.py", line 817, in test_ftp_errorself.assertEqual(raised.reason,
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^f"ftp error: {exception.args[0]}")
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^AssertionError: gaierror(7, 'No address associated with hostname') != 'ftp error: 500 OOPS: cannot change directory:/nonexistent'
----------------------------------------------------------------------
Ran 81 tests in 0.081s
FAILED (failures=1, skipped=3)