Bug report
Bug description:
I noticed this while working on GH-126890, as some odd macro redefinition warnings came up from the generated clinic file. As it turns out, clinic has a bug where using both @getter and @setter where the getter has a docstring doesn't work correctly. Quick reproducer:
/*[clinic input]
@getter
Test.test
My silly docstring
[clinic start generated code]*/
/*[clinic input]
@setter
Test.test
[clinic start generated code]*/
This results in generated code that looks like this:
PyDoc_STRVAR(Test_test__doc__,
"My silly docstring");
#define Test_test_HAS_DOCSTR
#if defined(Test_test_HAS_DOCSTR)
# define Test_test_DOCSTR Test_test__doc__
#else
# define Test_test_DOCSTR NULL
#endif
#if defined(TEST_TEST_GETSETDEF)
# undef TEST_TEST_GETSETDEF
# define TEST_TEST_GETSETDEF {"test", (getter)Test_test_get, (setter)Test_test_set, Test_test_DOCSTR},
#else
# define TEST_TEST_GETSETDEF {"test", (getter)Test_test_get, NULL, Test_test_DOCSTR},
#endif
static PyObject *
Test_test_get_impl(TestObj *self);
static PyObject *
Test_test_get(TestObj *self, void *Py_UNUSED(context))
{
return Test_test_get_impl(self);
}
#if defined(TEST_TEST_HAS_DOCSTR)
# define Test_test_DOCSTR Test_test__doc__
#else
# define Test_test_DOCSTR NULL
#endif
#if defined(TEST_TEST_GETSETDEF)
# undef TEST_TEST_GETSETDEF
# define TEST_TEST_GETSETDEF {"test", (getter)Test_test_get, (setter)Test_test_set, Test_test_DOCSTR},
#else
# define TEST_TEST_GETSETDEF {"test", NULL, (setter)Test_test_set, NULL},
#endifThere's two bugs here:
- The setter uses the wrong name for the
HAS_DOCSTR part; it uses TEST_TEST_HAS_DOCSTR while Test_test_HAS_DOCSTR, so the resulting docstring ends up being redefined to NULL.
- Even with that fixed, the docstring macro would get redefined anyway.
The simplest fix is to just drop the HAS_DOCSTR entirely, and just check whether the actual DOCSTR macro is defined or not, and then set it to NULL in that case (thanks @erlend-aasland for this suggestion).
I've created GH-127310 as a fix.
CPython versions tested on:
CPython main branch
Operating systems tested on:
Linux
Linked PRs
Bug report
Bug description:
I noticed this while working on GH-126890, as some odd macro redefinition warnings came up from the generated clinic file. As it turns out, clinic has a bug where using both
@getterand@setterwhere the getter has a docstring doesn't work correctly. Quick reproducer:This results in generated code that looks like this:
There's two bugs here:
HAS_DOCSTRpart; it usesTEST_TEST_HAS_DOCSTRwhileTest_test_HAS_DOCSTR, so the resulting docstring ends up being redefined to NULL.The simplest fix is to just drop the
HAS_DOCSTRentirely, and just check whether the actualDOCSTRmacro is defined or not, and then set it toNULLin that case (thanks @erlend-aasland for this suggestion).I've created GH-127310 as a fix.
CPython versions tested on:
CPython main branch
Operating systems tested on:
Linux
Linked PRs