René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Make Argument Clinic aware of possible UBSan failures ","articleBody":"# Feature or enhancement\r\n\r\nA huge part of the CPython code base uses Argument Clinic for generating methods for classes or standalone functions. Generally, methods are then used as follows:\r\n\r\n```C\r\nstatic PyMethodDef foo_methods[] = {\r\n    _FOOMODULE_FOO_BAR_METHODDEF\r\n    {NULL, NULL}\r\n};\r\n```\r\n\r\nwhere each `\u003cMODULE\u003e_\u003cCLASS\u003e_\u003cMETHOD\u003e_METHODDEF` is something like\r\n\r\n```C\r\nPyDoc_STRVAR(_foomodule_Foo_bar__doc__,\r\n\"bar($self, /)\\n\"\r\n\"--\\n\"\r\n\"\\n\");\r\n\r\n#define _FOOMODULE_FOO_BAR_METHODDEF \t\t\\\r\n    {\t\t\t\t\t\t\t\t\t\t\\\r\n        \"bar\", \t\t\t\t\t\t\t\t\\\r\n        (PyCFunction)_foomodule_Foo_bar, \t\\\r\n        METH_NOARGS, \t\t\t\t\t\t\\\r\n        _foomodule_Foo_bar__doc__           \\\r\n    },\r\n```\r\n\r\nWe also have\r\n\r\n```C\r\n// @file Modules/foomodule.c\r\n\r\n/*[clinic input]\r\n_foomodule.Foo.bar\r\n\r\n[clinic start generated code]*/\r\n\r\nstatic PyObject *\r\n_foomodule_Foo_bar_impl(FooObject *self)\r\n/*[clinic end generated code: output=... input=...]*/\r\n{\r\n\tPy_RETURN_NONE;\r\n}\r\n\r\n// @file Modules/clinic/foomodule.c.h\r\n\r\nstatic PyObject *\r\n_foomodule_Foo_bar_impl(FooObject *self);\r\n\r\nstatic PyObject *\r\n_foomodule_Foo_bar(FooObject *self, PyObject *Py_UNUSED(ignored))\r\n{\r\n    return _foomodule_Foo_bar_impl(self);\r\n}\r\n```\r\n\r\nNow, in order to fix #111178, we generally convert non-AC clinic methods by changing `FooObject *` to `PyObject *` and doing a pointer cast in the function body. See #111178 for a wider discussion. The issue is that we cannot easily do the same for AC-generated code since the prototype is part of the auto-generated block.\r\n\r\nWe can do it if we specify `self: self(type=\"PyObject *\")` instead of not specifying `self` in the AC input, but this is an overkill. Instead, AC should always generate two functions: an implementation function which takes a typed object (i.e., FooObject) and a public dispatcher which takes a PyObject *self and possibly unused arguments. Stated otherwise:\r\n\r\n```diff\r\nstatic PyObject *\r\n-_foomodule_Foo_bar(FooObject *self, PyObject *Py_UNUSED(ignored))\r\n+_foomodule_Foo_bar(PyObject *self, PyObject *Py_UNUSED(ignored))\r\n{\r\n-    return _foomodule_Foo_bar_impl(self);\r\n+    return _foomodule_Foo_bar_impl((FooObject *)self);\r\n}\r\n```\r\n\r\nI decided to open a separate issue to track this change orthogonally to the UBSan failures fixes that can be achieved without any AC changes. I hope that this will not deterioriate the performances of CPython too much.\r\n\r\n---\r\n\r\nNote: this does not only affect methods as above; this affects any *method* that is casted to `PyCFunction` or anything else with incompatible function pointer signatures.","author":{"url":"https://github.com/picnixz","@type":"Person","name":"picnixz"},"datePublished":"2024-12-22T15:02:59.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":4},"url":"https://github.com/128180/cpython/issues/128180"}