Skip to content

GH-155728: Untrack const tuples #156419

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
sergey-miryanov wants to merge 4 commits into
base: main
Choose a base branch
from
+20 −3
Open

GH-155728: Untrack const tuples #156419

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
ce997c2
Untrack tuples from _PyTuple_Concat and _PyTuple_Repeat
sergey-miryanov Aug 26, 2026
5064e61
Untrack from flowgraph optimizer
sergey-miryanov Aug 26, 2026
cbf8fed
Add tests
sergey-miryanov Aug 26, 2026
fee31fe
Add news
sergey-miryanov Aug 26, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 6 additions & 1 deletion Lib/test/test_tuple.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -319,10 +319,15 @@ def test_track_literals(self):
self._not_tracked_instantly((1,))
self._not_tracked_instantly((1, 2))
self._not_tracked_instantly((1, 2, "a"))
self._not_tracked_instantly((1, 2) * 5)
self._not_tracked_instantly((12, 10**10, 'a_' * 100))
self._not_tracked_instantly((object(),))

# Test for _PyTuple_Concat
self._not_tracked_instantly((1, 2) + (2, 3))

# Test for _PyTuple_Repeat
self._not_tracked_instantly((1, 2) * 5)

self._not_tracked(((1, x), y, (2, 3)))
self._not_tracked((1, 2, (None, True, False, ()), int))
self._not_tracked((object(), ()))
Expand Down
2 changes: 2 additions & 0 deletions Misc/NEWS.d/next/Core_and_Builtins/2026-08-26-19-28-49.gh-issue-155728.LJIPeZ.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,2 @@
The flowgraph optimizer now untracks tuples from the GC when possible. Patch
by Sergey Miryanov.
10 changes: 8 additions & 2 deletions Objects/tupleobject.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -593,7 +593,10 @@ _PyTuple_Concat(PyObject *aa, PyObject *bb)
dest[i] = Py_NewRef(v);
}

_PyObject_GC_TRACK(np);
if (_PyObject_GC_IS_TRACKED(aa) || _PyObject_GC_IS_TRACKED(bb))
{
_PyObject_GC_TRACK(np);
}
return (PyObject *)np;
}

Expand Down Expand Up @@ -642,7 +645,10 @@ _PyTuple_Repeat(PyObject *self, Py_ssize_t n)
_Py_memory_repeat((char *)np->ob_item, sizeof(PyObject *)*output_size,
sizeof(PyObject *)*input_size);
}
_PyObject_GC_TRACK(np);
if (_PyObject_GC_IS_TRACKED(self))
{
_PyObject_GC_TRACK(np);
}
return (PyObject *) np;
}

Expand Down
4 changes: 4 additions & 0 deletions Python/flowgraph.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,7 @@
#include "pycore_opcode_metadata.h" // OPCODE_HAS_ARG, etc
#include "pycore_pystate.h" // _PyInterpreterState_GET()
#include "pycore_stackref.h" // PyStackRef_AsPyObjectBorrow()
#include "pycore_tuple.h" // _PyTuple_MaybeUntrack()

#include <stdbool.h>

Expand Down Expand Up @@ -1563,6 +1564,7 @@ fold_tuple_of_constants(basicblock *bb, int i, PyObject *consts,
}
PyTuple_SET_ITEM(const_tuple, i, element);
}
_PyTuple_MaybeUntrack(const_tuple);

@sergey-miryanov sergey-miryanov Aug 26, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We could avoid a second pass there, just like in marshal, by checking each element. But I'm not sure we should do it, since it adds a bit of complexity.

I'm open to change this, if it is worth.

@efimov-mikhail efimov-mikhail Aug 27, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

IMO, we should do the same logic in each case, because we don't want to expose partially constructed tuples to the GC.
@markshannon, WDYT?

@sergey-miryanov sergey-miryanov Aug 27, 2026

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good point, will fix this.


nop_out(const_instrs, seq_size);
return instr_make_load_const(instr, const_tuple, consts, const_cache, consts_index);
Expand Down Expand Up @@ -1650,6 +1652,7 @@ fold_constant_seq_into_load_const(basicblock *bb, int i,
}
nop_out(&instr, 1);
}
_PyTuple_MaybeUntrack(newconst);
assert(consts_found == 0);

if (build_op == BUILD_SET) {
Expand Down Expand Up @@ -1737,6 +1740,7 @@ optimize_lists_and_sets(basicblock *bb, int i, int nextop,
}
PyTuple_SET_ITEM(const_result, i, element);
}
_PyTuple_MaybeUntrack(const_result);

if (instr->i_opcode == BUILD_SET) {
PyObject *frozenset = PyFrozenSet_New(const_result);
Expand Down
Loading