(original) (raw)

The proper fix for this in the code would likely break ABI compatibility (ie: not possible in python 2.7 or any other stable release).

Clang's UBSAN (undefined behavior sanitizer) has been flagging this one for a long time.

In Python 3 a double is used instead of long double since 2012 as I did some digging at the time: https://github.com/python/cpython/commit/e348c8d154cf6342c79d627ebfe89dfe9de23817

-gps

On Tue, Jan 30, 2018 at 10:59 AM Florian Weimer <fw@deneb.enyo.de> wrote:

I hope this is the right list for this kind of question. We recently
tried to build Python 2.6 with GCC 8, and ran into this issue:

<https://bugzilla.redhat.com/show\_bug.cgi?id=1540316>

Also quoting for context:

| PyInstance\_NewRaw contains this code:
|
| inst = PyObject\_GC\_New(PyInstanceObject, &PyInstance\_Type);
| if (inst == NULL) {
| Py\_DECREF(dict);
| return NULL;
| }
| inst->in\_weakreflist = NULL;
| Py\_INCREF(klass);
| inst->in\_class = (PyClassObject \*)klass;
| inst->in\_dict = dict;
| \_PyObject\_GC\_TRACK(inst);
|
| \_PyObject\_GC\_TRACK expands to:
|
| #define \_PyObject\_GC\_TRACK(o) do { \\
| PyGC\_Head \*g = \_Py\_AS\_GC(o); \\
| if (g->gc.gc\_refs != \_PyGC\_REFS\_UNTRACKED) \\
| Py\_FatalError("GC object already tracked"); \\
| …
|
| Via:
|
| #define \_Py\_AS\_GC(o) ((PyGC\_Head \*)(o)-1)
|
| We get to this:
|
| /\* GC information is stored BEFORE the object structure. \*/
| typedef union \_gc\_head {
| struct {
| union \_gc\_head \*gc\_next;
| union \_gc\_head \*gc\_prev;
| Py\_ssize\_t gc\_refs;
| } gc;
| long double dummy; /\* force worst-case alignment \*/
| } PyGC\_Head;
|
| PyGC\_Head has 16-byte alignment. The net result is that
|
| \_PyObject\_GC\_TRACK(inst);
|
| promises to the compiler that inst is properly aligned for the
| PyGC\_Head type, but it is not: PyObject\_GC\_New returns a pointer which
| is only 8-byte-aligned.
|
| Objects/obmalloc.c contains this:
|
| /\*
| \* Alignment of addresses returned to the user. 8-bytes alignment works
| \* on most current architectures (with 32-bit or 64-bit address busses).
| \* The alignment value is also used for grouping small requests in size
| \* classes spaced ALIGNMENT bytes apart.
| \*
| \* You shouldn't change this unless you know what you are doing.
| \*/
| #define ALIGNMENT 8 /\* must be 2^N \*/
| #define ALIGNMENT\_SHIFT 3
| #define ALIGNMENT\_MASK (ALIGNMENT - 1)
|
| So either the allocator alignment needs to be increased, or the
| PyGC\_Head alignment needs to be decreased.

Is this a known issue? As far as I can see, it has not been fixed on
the 2.7 branch.

(Store merging is a relatively new GCC feature. Among other things,
this means that on x86-64, for sufficiently aligned pointers, vector
instructions are used to update multiple struct fields at once. These
vector instructions can trigger alignment traps, similar to what
happens on some other architectures for scalars.)
\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_
Python-Dev mailing list
Python-Dev@python.org
https://mail.python.org/mailman/listinfo/python-dev
Unsubscribe: https://mail.python.org/mailman/options/python-dev/greg%40krypto.org