Differences between revisions 48 and 49
Revision 48 as of 2008-08-15 19:07:32
Size: 3177
Editor: c-67-190-72-101
Comment:
Revision 49 as of 2008-08-15 19:35:35
Size: 3206
Editor: c-67-190-72-101
Comment:
Deletions are marked like this. Additions are marked like this.
Line 13: Line 13:
* test_doctest - test_doctest2.txt, which is run by test_DocFileSuite, has some random Unicode characters, but no encoding declared (looks utf8). Does this rely on an implicit utf8 encoding? * test_doctest - test_doctest2.txt, which is run by test_DocFileSuite, includes the string u'\xef\xbb\xbf\xd0\x89\xd0\x8a\xd0\x88\xd0\x81\xd0\x82'. The encoding is declared to be utf8 in the call to doctest.DocFileSuite

To Fix

  • test_builtin
    • test_input_and_raw_input - del sys.stdin does nothing
    • test_cmp - Jython can compare objects with cycles in them, at least for this test.
    • test_general_eval - Jython can use arbitrary maps for globals; this was discussed for CPython. Not in current 2.6, perhaps because it touched a lot of code or issues with performance (see http://bugs.python.org/issue1402289)
    • test_execfile - on the other hand, Jython does not even validate that the locals or globals are in fact mapping objects
  • test_doctest - test_doctest2.txt, which is run by test_DocFileSuite, includes the string u'xefxbbxbfxd0x89xd0x8axd0x88xd0x81xd0x82'. The encoding is declared to be utf8 in the call to doctest.DocFileSuite
  • test_generators
    • parser probs where we don't reject bad syntax
    • ill-behaved generator doesn't finalize upon GC
    • closure setup in visitLambda and visitGeneratorExp does not interact properly with a yield point, resulting in bad bytecode
  • test_inspect
    • test_excluding_predicates - sys.exit, etc., are not really attributes of a module
    • test_wrapped_decorator - see below in the Parser section
  • test_site - del sys.setdefaultencoding does nothing
  • test_struct - needs new 2.5 struct.Struct type and pack_into/unpack_from
  • test_unicode
    • test_codecs_errors - raises wrong error for decode to int of unicode string
    • test_codecs_idna - no idna encoding (which needs unicodedata)
  • test_univnewlines - test_doctest failures don't restore sys.stdin -- causes to skip (it normally passes)
  • test_with - 1 failure, testNestedSingleStatements (__main__.ExceptionalTestCase), this is a tricky one

Flaky/arch?

  • test_cpickle
  • test_descr
  • test_socket - historically this is the case on SO_RCVBUF, SO_SNDBUF
  • test_scope
  • test_tuple - sometimes OutOfMemoryErrors during regrtest (when creating a set from a large list of small tuples). maybe not the tests fault, but the fact that the regrtest progressively uses more and more memory as it goes: http://pylonshq.com/pasties/917

Parser

  • test_compiler - one line functions are not working compiler.parse (causing AttributeError: 'NoneType' object has no attribute 'suite')
  • test_doctest - in test_pdb_set_trace, removing # doctest: +NORMALIZE_WHITESPACE allows it to parse
  • test_decorators - compile(..., "eval") not working for test_dbcheck
  • test_doctest - in addition to the unicode problem, there is a parsing problem with newlines + comment -- the following fails: (try: pass<newline>except: pass<newlne>#comment)
  • test_syntax - lots of work needed to make exceptions look more like CPython
  • test_inspect.test_wrapped_decorator, the line number needs to be set for the wrapped function at the def, not at the decorators stacked above. right now node.getLine() in CodeCompiler#visitFunctionDef must be reporting where the function def starts in the AST (which includes of course the decorators)

Development

  • test__rawffi (for ctypes, ported from PyPy)
  • test_multibytecodec (implies test_multibytecodec_support)
  • test_unicodedata - need to implement unicodedata!

UpgradeTo25CPythonLib (last edited 2010-05-25 03:25:59 by vpn3-145126)