
I've spent an aggravating little while running CPython's regrtest with a recent pypy-c (default options, so boehm and no fancy stuff). After a bit of trial and error, this was the result: 69 tests failed: GC Warning: Finalization cycle involving d66c120 GC Warning: Finalization cycle involving a8e06c0 GC Warning: Finalization cycle involving d286d08 GC Warning: Finalization cycle involving d180780 GC Warning: Finalization cycle involving d12a108 test___all__ test_anydbm test_calendar test_cfgparser test_class test_codecencodings_cn test_codecencodings_hk test_codecencodings_jp test_codecencodings_kr test_codecencodings_tw test_coercion test_commands test_compare test_compiler test_cookielib test_cpickle test_decimal test_descr test_descrtut test_dis test_distutils test_doctest2 test_dumbdbm test_extcall test_file test_filecmp test_fileinput test_frozen test_gc test_getargs test_gettext test_glob test_inspect test_iter test_mailbox test_marshal test_mhlib test_mimetools test_minidom test_multibytecodec test_new test_os test_peepholer test_pep292 test_pickle test_pickletools test_posixpath test_profile test_pyclbr test_random test_repr test_scope test_sort test_strftime test_stringprep test_strptime test_structseq test_tarfile test_tempfile test_time test_traceback test_transformer test_ucn test_unicodedata test_weakref test_whichdb test_xmlrpc test_xpickle test_zipfile 97 tests skipped: test__locale test_aepack test_al test_applesingle test_asynchat test_audioop test_bsddb test_bsddb185 test_bsddb3 test_bz2 test_capi test_cd test_cgi test_cl test_codecmaps_cn test_codecmaps_hk test_codecmaps_jp test_codecmaps_kr test_codecmaps_tw test_crypt test_csv test_curses test_dbm test_dl test_doctest test_email test_email_codecs test_fcntl test_fork1 test_gdbm test_getargs2 test_gl test_grp test_gzip test_hotshot test_httplib test_imageop test_imaplib test_imgfile test_imp test_ioctl test_largefile test_linuxaudiodev test_locale test_logging test_macfs test_macostools test_mimetypes test_mmap test_nis test_normalization test_openpty test_ossaudiodev test_pep277 test_plistlib test_poll test_popen test_popen2 test_pty test_pwd test_pyexpat test_queue test_regex test_resource test_rgbimg test_robotparser test_sax test_scriptpackages test_select test_signal test_site test_socket test_socket_ssl test_socketserver test_strop test_subprocess test_sunaudiodev test_sundry test_symtable test_tcl test_thread test_threaded_import test_threadedtempfile test_threading test_threading_local test_threadsignals test_timeout test_timing test_unicode_file test_urllib test_urllib2 test_urllib2net test_urllibnet test_winreg test_winsound test_zipimport test_zlib Segmentation fault I had to run with '-x test_datetime test_format test_shutil test_shelve test_str test_string test_unicode test_userstring' in the end. test_datetime, test_shutil and test_format kill the build. the string tests fail in ways that suggest that the boehm gc really didn't expect the usage patterns they use. I don't remember why I skipped test_shelve now :) I count around 250 test_* files in CPython's test suite, so we actually pass around 40% of them, and a lot of the skips and plenty of the fails are missing (parts of) extension modules, so we're not really in *that* bad a shape. But fixing the crashers would be nice. We need to get some automated version of this set up. Cheers, mwh -- ARTHUR: Why should he want to know where his towel is? FORD: Everybody should know where his towel is. ARTHUR: I think your head's come undone. -- The Hitch-Hikers Guide to the Galaxy, Episode 7

Michael Hudson <mwh@python.net> writes:
I've spent an aggravating little while running CPython's regrtest with a recent pypy-c (default options, so boehm and no fancy stuff).
After a bit of trial and error, this was the result:
69 tests failed: 97 tests skipped:
In some good news, the stacklessgc build passes exactly one fewer test (test_builtin, known issue) and doesn't even do this:
Segmentation fault
and the end :) Though it does exit with a rather strange ImportError: No module named _socket Haven't dug yet. On this evidence, I think I'm inclined to say that framework GC builds are at least as solid as the Boehm builds. Cheers, mwh -- C is not clean -- the language has _many_ gotchas and traps, and although its semantics are _simple_ in some sense, it is not any cleaner than the assembly-language design it is based on. -- Erik Naggum, comp.lang.lisp
participants (1)
-
Michael Hudson