bpo-43950: ensure source_line is present when specializing the traceback

[isidentical]

There seems to be different cases where the _PyDisplay_Line() function returns 0 without displaying the actual line but also without having any errors raised.

cpython/Python/traceback.c

Lines 414 to 420 in 4512848

	/* use the right encoding to decode the file as unicode */
	fd = PyObject_AsFileDescriptor(binary);
	if (fd < 0) {
	Py_DECREF(io);
	Py_DECREF(binary);
	return 0;
	}

cpython/Python/traceback.c

Lines 436 to 446 in 4512848

	if (fob == NULL) {
	PyErr_Clear();
	res = _PyObject_CallMethodIdNoArgs(binary, &PyId_close);
	Py_DECREF(binary);
	if (res)
	Py_DECREF(res);
	else
	PyErr_Clear();
	return 0;
	}

This patch ensures that the line is retrieved and displayed.

https://bugs.python.org/issue43950

[sweeneyde]

Would it be better to fix _Py_DisplaySourceLine() instead?

[isidentical]

I am not sure how would the fix look beside introducing a new return code for cases that did not result with printing the line but also without an error which seems more complicated.

…

On Fri, Jul 23, 2021, 10:45 PM Dennis Sweeney ***@***.***> wrote: Would it be better to fix _Py_DisplaySourceLine() instead? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#27313 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ALJKHQPWWFKTA45LSTO5FS3TZHBGFANCNFSM5A4RI23Q> .

[pablogsal]

Can we add a regression test for this? (Like the one on the BPO issue?)

[bedevere-bot]

🤖 New build scheduled with the buildbot fleet by @isidentical for commit 8db4e0c 🤖

If you want to schedule another build, you need to add the ":hammer: test-with-buildbots" label again.

[isidentical]

It seems like refleaks bots also fail on the main too, so that can be ignored.