`exec(<string>, closure=<non-None>)` failed assertion

[bswck]

Crash report

What happened?

Minimal repro:

print("Reproducing!")
exec("", closure=object())

Fails with

Reproducing!
python: Python/ast.c:1047: _PyAST_Validate: Assertion `!PyErr_Occurred()' failed.
Aborted (core dumped)

Note the error occurs not because the passed closure= is incorrect (it should be a tuple of cell objects), but because closure= is not allowed when string is the source.

The reason for that to happen is a missing goto error jump:

cpython/Python/bltinmodule.c

Lines 1175 to 1178 in 2a0256f

	if (closure != NULL) {
	PyErr_SetString(PyExc_TypeError,
	"closure can only be used when source is a code object");
	}

causing the rest of the code to execute and reach _PyAST_Validate.

In the PR, I included the missing test case.

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Output from running 'python -VV' on the command line:

Python 3.14.0a5+ (heads/main-dirty:f9a7d41bacb, Feb 13 2025, 04:52:15) [GCC 11.4.0]

Linked PRs

• gh-130070: Fix exec(<string>, closure=<non-None>) unexpected path #130071
• [3.13] gh-130070: Fix exec(<string>, closure=<non-None>) unexpected path (GH-130071) #132627

[StanFromIreland]

Unable to reproduce in 3.13 or main branch build with the supplied code:

Reproducing!
Traceback (most recent call last):
  File "<python-input-0>", line 2, in <module>
    exec("", closure=object())
    ~~~~^^^^^^^^^^^^^^^^^^^^^^
TypeError: closure can only be used when source is a code object

[bswck]

Oh! This isn't a segfault. It says "aborted".

[bswck]

@tomasr8 Should the label be changed to type-bug? It's not a hard crash, just a subtle bug, even though a core dump is produced.

[encukou]

It crashes in debug builds.