)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"5f3501dbd8582547b047178e073d96c272b62126","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"fdd36261_d36af68d","updated":"2025-04-11 20:32:57.000000000","message":"Oh, hmm...","commit_id":"06622b57cb769523dbda7f296223f3619fd2745e"}],"swift/common/utils/checksum.py":[{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"f97132a661a55fc734d2ac805ceeeaad9c6056b9","unresolved":true,"context_lines":[{"line_number":99,"context_line":"    #"},{"line_number":100,"context_line":"    # i.e., ISA-L is consistently 3-5x faster than kernel sockets,"},{"line_number":101,"context_line":"    # which is still \u003e1000x faster than a naive python implementation."},{"line_number":102,"context_line":"    selected \u003d crc32c_isal or crc32c_kern or None"},{"line_number":103,"context_line":"    if not selected:"},{"line_number":104,"context_line":"        raise NotImplementedError(\u0027no crc32c implementation, install isal\u0027)"},{"line_number":105,"context_line":"    return selected"}],"source_content_type":"text/x-python","patch_set":1,"id":"858cfd48_294fecd5","line":102,"range":{"start_line":102,"start_character":42,"end_line":102,"end_character":49},"updated":"2025-04-11 15:31:39.000000000","message":"Not strictly necessary, but probably worth it to keep it explicit.","commit_id":"06622b57cb769523dbda7f296223f3619fd2745e"}],"test/unit/__init__.py":[{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"f97132a661a55fc734d2ac805ceeeaad9c6056b9","unresolved":true,"context_lines":[{"line_number":1091,"context_line":"        try:"},{"line_number":1092,"context_line":"            checksum.crc32c()"},{"line_number":1093,"context_line":"        except NotImplementedError as e:"},{"line_number":1094,"context_line":"            raise SkipTest(str(e))"},{"line_number":1095,"context_line":"        return func(*args, **kwargs)"},{"line_number":1096,"context_line":"    return wrapper"},{"line_number":1097,"context_line":""}],"source_content_type":"text/x-python","patch_set":1,"id":"2a31a8d0_048f2af1","line":1094,"updated":"2025-04-11 15:31:39.000000000","message":"Right, so this closely mirrors the behavior in the server at `_get_checksum_hasher`","commit_id":"06622b57cb769523dbda7f296223f3619fd2745e"}],"test/unit/common/utils/test_checksum.py":[{"author":{"_account_id":15343,"name":"Tim Burke","email":"tburke@nvidia.com","username":"tburke"},"change_message_id":"5f3501dbd8582547b047178e073d96c272b62126","unresolved":true,"context_lines":[{"line_number":133,"context_line":"        # Check preferences -- ISA-L always wins"},{"line_number":134,"context_line":"        self.assertIs(checksum._select_crc32c_impl(), checksum.crc32c_isal)"},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"    def test_equivalence(self):"},{"line_number":137,"context_line":"        data \u003d os.urandom(64 * 1024)"},{"line_number":138,"context_line":"        ref_crc \u003d checksum.crc32c_ref(data)"},{"line_number":139,"context_line":"        self.assertEqual(checksum.crc32c(data).crc, ref_crc)"}],"source_content_type":"text/x-python","patch_set":1,"id":"5cd02be4_c8cf3c5b","side":"PARENT","line":136,"updated":"2025-04-11 20:32:57.000000000","message":"I\u0027m not sure about dropping this test -- there seem to be just enough complexities to those kernel sockets and ctypes bindings that I think there\u0027s still value in having some spot-checking of equivalence beyond just the check value... if anything, I think it might be worth _expanding_ this to ensure that `crc(data, crc(data)) \u003d\u003d crc(data * 2)`\n\nSpeaking of complexities, I should re-investigate that comment about\n```\n        # for some reason, despite us specifying that restype is uint,\n        # it can come back signed??\n```","commit_id":"bfd0722e92a61c0f122ddc0c172a3be3876e2b27"},{"author":{"_account_id":7847,"name":"Alistair Coles","email":"alistairncoles@gmail.com","username":"acoles"},"change_message_id":"db70be23cd76b02513ca962b1875ff5d43a795e8","unresolved":true,"context_lines":[{"line_number":133,"context_line":"        # Check preferences -- ISA-L always wins"},{"line_number":134,"context_line":"        self.assertIs(checksum._select_crc32c_impl(), checksum.crc32c_isal)"},{"line_number":135,"context_line":""},{"line_number":136,"context_line":"    def test_equivalence(self):"},{"line_number":137,"context_line":"        data \u003d os.urandom(64 * 1024)"},{"line_number":138,"context_line":"        ref_crc \u003d checksum.crc32c_ref(data)"},{"line_number":139,"context_line":"        self.assertEqual(checksum.crc32c(data).crc, ref_crc)"}],"source_content_type":"text/x-python","patch_set":1,"id":"32605047_25e7f1d0","side":"PARENT","line":136,"in_reply_to":"5cd02be4_c8cf3c5b","updated":"2025-04-14 08:47:09.000000000","message":"yep, I was reflecting on this and wondered if we might just move the ref impl\u0027s to the test module to keep things honest. Or at least compare kern to isal","commit_id":"bfd0722e92a61c0f122ddc0c172a3be3876e2b27"}]}