Author: Armin Rigo <[email protected]>
Branch:
Changeset: r78375:811532eaa04d
Date: 2015-07-01 11:24 +0200
http://bitbucket.org/pypy/pypy/changeset/811532eaa04d/
Log: Skip the other test with a comment.
diff --git a/pypy/module/imp/test/test_import.py
b/pypy/module/imp/test/test_import.py
--- a/pypy/module/imp/test/test_import.py
+++ b/pypy/module/imp/test/test_import.py
@@ -1217,7 +1217,16 @@
finally:
sys.path_hooks.pop()
- def test_meta_path_import_star_1(self):
+ def test_meta_path_import_error_1(self):
+ # as far as I can tell, the problem is that in CPython, if you
+ # use an import hook that doesn't update sys.modules, then the
+ # import succeeds; but at the same time, you can have the same
+ # result without an import hook (see test_del_from_sys_modules)
+ # and then the import fails. This looks like even more mess
+ # to replicate, so we ignore it until someone really hits this
+ # case...
+ skip("looks like an inconsistency in CPython")
+
class ImportHook(object):
def find_module(self, fullname, path=None):
assert not fullname.endswith('*')
@@ -1225,12 +1234,13 @@
return self
def load_module(self, fullname):
assert fullname == 'meta_path_pseudo_module'
+ # we "forget" to update sys.modules
return new.module('meta_path_pseudo_module')
import sys, new
sys.meta_path.append(ImportHook())
try:
- exec "from meta_path_pseudo_module import *" in {}
+ import meta_path_pseudo_module
finally:
sys.meta_path.pop()
_______________________________________________
pypy-commit mailing list
[email protected]
https://mail.python.org/mailman/listinfo/pypy-commit
