In order to upgrate the tests for ctypes, following changes are required:
 - binascii.hexlify should accept memoryview() objects
 - the ctypes character buffer should accept assignment of memoryview() objects using their "raw" property
Then we can backport the Py3 ctypes tests to Py2.

Patches attached:
 - partial backport of Modules/binascii.c
 - partial backport of Modules/_ctypes/_ctypes.c (preserving 2.3 compat.)
 - update ctypes tests (buffer -> memoryview)

Does binascii still work with array.array objects? Also, there should be additional tests for compatibility of binascii with memoryview objects.

Additional tests for binascii.

Please be careful with the coding style. Stuff like:
+			  if (_PyString_Resize(&rv, 2*out_len) < 0) \
+			    { Py_DECREF(rv); PyBuffer_Release(&pin); return NULL; } \
should be spread out on several lines.
Otherwise, the binascii patch looks good.

I've now committed the binascii patch + tests to trunk, and merged the tests into py3k.
Note: binascii_a2b_hqx() still uses the "t#" argument specifier in py3k as well as in trunk, this would deserve a separate patch.

Patch for the last "t#" format string in binascii module.
And provide additional tests.
PS: I removed this comment. IMHO, it is a wrong assertion:
        "# The hqx test is in test_binhex.py"

I don't think the bytes -> bytearray changes are useful, e.g.:
-        fillers = ""
+        fillers = bytearray()
As for:
+            try:
+                f(empty)
+            except SystemError, err:
+                self.fail("{}({!r}) raises SystemError: {}".format(func, empty, err))
+            except Exception, err:
+                self.fail("{}({!r}) raises {!r}".format(func, empty, err))
The "except SystemError" is pointless, since "except Exception" will catch SystemError anyway.

Changed.

Removed no-op str("...") conversions.

When the following snippet is taken:
	if (_PyString_Resize(&rv,
			   (bin_data -
			    (unsigned char *)PyString_AS_STRING(rv))) < 0) {
		PyBuffer_Release(&pascii);
		Py_DECREF(rv);
		rv = NULL;
	}
pascii will get released a second time at the end of function. The rest is fine, I'm gonna fix it myself.

The a2b_hqx() patch was committed in r77528 (trunk) and r77529 (py3k). Thanks again!

Fixed with r79288 and r79295.

resolution: fixed
stage: patch review -> resolved

messages: + msg97877

messages: + msg97872

messages: + msg97867

messages: + msg97769

assignee: theller
components: + ctypes

nosy: + theller
messages: + msg97766
stage: needs patch -> patch review