Fix test_pickle, by reverting the string opcodes (S, T, U) to returning

strings, in Latin-1.  Bytes are once more pickled through bytes.__reduce__,
but now it returns "latin-1" as the second parameter.

Unfortunately this breaks datetime pickling.  I'll have to investigate
further; reverting Martin's changes doesn't seem to help.
This commit is contained in:
Guido van Rossum 2007-07-19 22:19:35 +00:00
parent 00058aa28c
commit f93254d299
2 changed files with 6 additions and 22 deletions

View file

@ -2724,13 +2724,11 @@ PyDoc_STRVAR(reduce_doc, "Return state information for pickling.");
static PyObject *
bytes_reduce(PyBytesObject *self)
{
/* XXX: This currently returns a Py_UNICODE-widened string
in the tuple which is completely useless. Pickle stopped
using it for that reason. */
return Py_BuildValue("(O(s#))",
return Py_BuildValue("(O(s#s))",
self->ob_type,
self->ob_bytes == NULL ? "" : self->ob_bytes,
self->ob_size);
self->ob_size,
"latin-1");
}
static PySequenceMethods bytes_as_sequence = {