View source
py3k
  • Contributors
    1. Loading...
Author Commit Message Date Builds
250+ commits behind default.
Romain Guillebert
Adapt test_proxy_iter.py to Python 3's syntax
Antonio Cuni
bah
Antonio Cuni
move the decoding of valuefmt later, else the annotation cannot determine that the argument to get_operation_class (which is memo-specialized) is a PBC
Antonio Cuni
we no longer need to pass unicode literals to operationerrfmt, because now the exception messages are unicode anyway
Antonio Cuni
make all the messages computed by operationerrfmt as unicode, even if the source message is a byte string
Antonio Cuni
backout 5226ba3dc28d, I'll try another approach for unicode exception messages
Antonio Cuni
backout 81cd817ffe62, I'll try another approach for unicode exception messages
Antonio Cuni
bah, there are two levels of memo/cache, and we need to specialize for unicode formatting strings on both levels
Antonio Cuni
the exception message is an unicode now
Antonio Cuni
we call unicode_w now
Antonio Cuni
we need this now
Antonio Cuni
fix test_destructor by using the actual array module instead of faking it
Antonio Cuni
fix test_argument.test_unicode_keywords by finally using unicode to store the exception message
Antonio Cuni
add the possibility to have unicode error messages with operrfmt
Antonio Cuni
fix -A support when the source code contains unicode chars. Fix test_unicode_keywords to pass on CPython3 (still fails on pypy)
Antonio Cuni
add a comment
Antonio Cuni
fix test_strutil now that it accepts unicode
Antonio Cuni
hg merge default
Antonio Cuni
hg merge default
Antonio Cuni
bah, we need to recursively convert strings to unicode also inside constant tuples
Antonio Cuni
bah; rpython+unicode is hard
Antonio Cuni
bah, we cannot use unicode.lower in rpython. Work-around that
Antonio Cuni
rever part of the previous checkins and use @with_unicode_literals instead of puttin u'' everywhere, to avoid too much divergence from default
Antonio Cuni
alternate format is now valid also for complex
Antonio Cuni
fix parsing of complex numbers by using unicode instead of strings
Antonio Cuni
add a decorator which allows to selectively use the equivalent of "from __future__ import unicode_literals" on a per-function basis
Antonio Cuni
bah
Antonio Cuni
latin1 is not rpython, latin-1 is
Antonio Cuni
fix int(bytearray(...))
Antonio Cuni
add a test to check the point of the whole refactoring, i.e. that non-ASCII chars are preserved in the exception message
Antonio Cuni
hg merge default
Antonio Cuni
we no longer have objects of type w_str, kill the corresponding code paths
Antonio Cuni
try hard to give good error messages when we are unable to convert a string to int() or float(). To do so, we do the parsing directly in unicode instead of trying to convert to ASCII and do the parsing there.
Antonio Cuni
hg merge default
Antonio Cuni
merge heads
Antonio Cuni
don't crash if we try to wrap a non-ascii byte string; this might still happen because e.g. exception messages are not unicode yet
Antonio Cuni
more differences in error messages
Antonio Cuni
20 nested blocks is a limitation of cpython; skip this test else it hangs: this is a partial port of rev b3bb9ebd9346 for 2.7
Antonio Cuni
pypy has different error messages
mattip
merge to head
mattip
fix skip
Antonio Cuni
don't import AppTestTypeObject directly, else the class is collected as well and we run the test twice
Antonio Cuni
exceptions are no longer in the exceptions module now
mattip
back out latest changesets, retain python2.x syntax
mattip
py3k syntax
mattip
whoops (antocuni)
mattip
fix for python 2.6
mattip
adapt syntax to py3k
Antonio Cuni
import the types directly from __exception__ instead of builtins. This makes a difference because the conftest overrides the AssertionError in the builtins, but we want to test our own
Antonio Cuni
the 'exceptions' module is gone in Python3, and exceptions are now directly in builtins. However, in PyPy we cannot simply move them to builtins, because they are needed in the early bootstrap of the space, before builtins is initialized. So, we keep them in a separate module (renamed to '__exceptions__' because it's an internal implementation detail only) but we pretend that its __module__ is 'builtins'. This approach has the extra bonus that it minimizes the divergence from default.
mattip
py3k syntax fix
Antonio Cuni
use funcargs instead of globals for demo_script and crashing_demo_script; it's needed because setup_class changes the cwd
Antonio Cuni
bah, typo
soareschen
Add alternate float formatting styles to new-style formatting to PyPy Py3k. Based on Python issue #7094. Also switch complex number formatting implementation to use rfloat.double_to_string instead of formatd. Contributors: Soares Chen <soares.chen@gmail.com> Tomasz Rybak <tomasz.rybak@post.pl>
Antonio Cuni
make autopath and our pytest.py compatible with python3: this is needed to run -A tests with the compiled pypy-c
Antonio Cuni
py3k-ify
Antonio Cuni
kill the old python2 io encoding stuff
Antonio Cuni
hg merge default
Antonio Cuni
pyc files are now stored in __pycache__, fish it from there
Antonio Cuni
I don't know why 'import site failed' was added to the expected stdout. Revert the test to what it is on default, and it passes again
Antonio Cuni
hg merge app_main-refactor again
Antonio Cuni
merge the app_main-refactor branch; tests are stil broken because we can't import autopath and in general the pypy tree in python3
Antonio Cuni
kill lines which were wrongly put in by a bad merge in rev a9fa4e709310
Antonio Cuni
fix test_print_info by making app_main importable on python2
Antonio Cuni
yet another way to set the correct PYTHONPATH, which works also for other test classes
Antonio Cuni
having the correct PYTHONPATH is required by more tests, so move the logic in _spawn
Antonio Cuni
bah, actually fix test_options_i_m: we need to ensure to have the correct pypy package in PYTHONPATH
Antonio Cuni
make pypy importable in python3; this fixes test_app_main.test_options_i_m, which try to import pypy.translator.goal.test2.mymodule
Antonio Cuni
try to simulate the very same conditions we get when running app_main on pypy: in particular, there are no sys.std{in,out,err} yet
Antonio Cuni
we cannot directly import/execute pypy.module.sys.state.getinitalpath, because it's a python2 file. But the good news is that we don't actually need it: for the tests, it is enough to have a minimal path, so we just compute it (in particular, we need to be able to import _pypy_interact)
Antonio Cuni
make sure to init sys.std{in,out,err} only if they are not initialized yet. This fixes the first few tests in test_app_main, which rely on stdout to be a StringIO()
Antonio Cuni
str vs bytes
Antonio Cuni
we print lots of [platform:execute] lines now, so we need to be a bit more careful about what to match with the regexp
Antonio Cuni
bah: we want to call bytes([n]), not bytes(n). space.call does the latter, because it receives a list of arguments, while space.call_function reveives *args
Antonio Cuni
py3k_skip tests for ropes. We are not interested in those right now
Antonio Cuni
kill again rangelist tests, which were resurrected by mistake by the last emrge
Antonio Cuni
hg merge default
Antonio Cuni
list this missing test file
Antonio Cuni
s/thread/_thread
Antonio Cuni
make sure to import a module which does not import anything, else the counter might be >1
Antonio Cuni
speedup tests by passing -S to py.py
Antonio Cuni
baaaah. Veeery subtle difference wrt python2
Antonio Cuni
we need to explicitly save exceptions now
Antonio Cuni
we need to explicitly save exceptions now
Antonio Cuni
isinstance(x, str) is not rpython, so it's nonsense to have it in the fake objspace. This fixes module/posix/test/test_ztranslation
Antonio Cuni
increase pexpect's maxread, because now we print a bit more [platform:execute] lines
Antonio Cuni
itertools.izip has gone in py3k
Antonio Cuni
bytes vs unicode and print() function
Antonio Cuni
subtle py2 vs py3 difference
Antonio Cuni
py3k{ify,skip} some more tests
Antonio Cuni
py3k-ify
Antonio Cuni
py3k_skip these tests for now
Antonio Cuni
improve the fake objspace to make checkmodule('_random') happy
Antonio Cuni
kill long literals
Antonio Cuni
py3k_skip one test, fix the other
Antonio Cuni
py3k-ify. Some tests fail for other reasons though (e.g., missing _socket.dup)
Antonio Cuni
py3k-ify, and pass -S to py.py to speed up tests
Amaury Forgeot d'Arc
Some translation fixes
Amaury Forgeot d'Arc
Many fixes in rope objects for py3k
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
Fix various test failures
Antonio Cuni
merge
Antonio Cuni
hg merge default
Antonio Cuni
str vs bytes
Antonio Cuni
_md5 is no longer there
Antonio Cuni
failing test&fix. I think that bufferstr_w is what we want, not str_w; it makes a difference e.g. for bytearray and memoryview
Antonio Cuni
make it possible to do int(bytearray(...)). This fixes test_bytearrayobject.test_int
Antonio Cuni
yet another place where to update the version number
Amaury Forgeot d'Arc
Fixes after merge
Amaury Forgeot d'Arc
Merge glitch
Amaury Forgeot d'Arc
hg merge default. Note: the step-one-xrange optimization was not ported.
Amaury Forgeot d'Arc
Advertise the correct Python version in our Python.h
Antonio Cuni
explicitly save exception values
Antonio Cuni
merge heads
Antonio Cuni
exception variables are cleared as soon as we exit the except: block now, so we must explicitly save the values
Armin Rigo
Dummy merge of default at a7e99ee14924, which is just after the previous merge of default, killing lib-python/3.2. In this merge I don't actually kill lib-python/3.2, of course.
Antonio Cuni
properly implement POP_EXCEPT as a block, instead of popping the last_exception from the stack directly in the implementation of the opcode. This fixes test_pop_exception_value. Took two days to hunt&fix :-(
Antonio Cuni
hg merge default (again)
Antonio Cuni
hg merge default
Antonio Cuni
generate the very same code as cpython for try/except blocks. In particular, make sure to delete the local variable which contains the exception after we exit the except: block
Antonio Cuni
add a failing test
Antonio Cuni
py3k-ify
Antonio Cuni
hg merge default
Antonio Cuni
some of these tests require array. It works on default because the array type is faked, but here we need the actualy py3k one, not the faked py2 array
Ronny Pfannschmidt
merge stdlib-unification/py3k
Ronny Pfannschmidt
close before merge
Ronny Pfannschmidt
merge from py3k
Antonio Cuni
py3k-ify by killing the u'' string prefix
Antonio Cuni
the re.UNICODE flags is by default now. Disable it explicitly for the test
Ronny Pfannschmidt
merge from stdlib-unification
Ronny Pfannschmidt
merge from py3k
Antonio Cuni
py3k-ify by killing the u'' string prefix
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
Adapt many cpyext tests to python3.
Amaury Forgeot d'Arc
cpyext: fixes in test_unicodeobject.py
Amaury Forgeot d'Arc
cpyext: Update _sre.c used in tests with a recent 3.2 version, and convert some print statements
Amaury Forgeot d'Arc
cpyext: copy the 3.2 version of arraymodule.c
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
To fake a py3k generator on top of pyhon2, it's necessary to rename .next() to .__next__(). This fixes space.wrap((x for x in range(10)))
Amaury Forgeot d'Arc
Various fixes to cpyext, until the test suite executes all tests. (103 failures, 227 passed)
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
update cpyext src directory with CPython 3.2 source code
Amaury Forgeot d'Arc
hg merge default
Antonio Cuni
py3k_skip for now the tests which fails because they need to handle unicode and longs now
Antonio Cuni
py3k_skip for now the tests which fails because they need to handle unicode and longs now
Antonio Cuni
oldstyle classes are gone, kill this test
Antonio Cuni
comment out an assert which is no longer valid in py3k
Antonio Cuni
py3k_skip these two tests
Antonio Cuni
py3k_skip this test while waiting for a UnicodeDictStrategy
Antonio Cuni
py3k-ify
Antonio Cuni
we no longer have oldstyle classes, kill the test
Antonio Cuni
allow lookup of str subclasses in dictproxies. Comment out py2-only checks
Antonio Cuni
py3k_skip two tests which fails right now because of missing strategies
Antonio Cuni
py3k-ify, __getslice__ vs __getitem__
Antonio Cuni
classic classes are gone in py3k
Antonio Cuni
py3k-ify: maxint has gone, and __getitem__ gets a slice() object now. The test was previously written with __getslice__ but got changed in a parent revision
Antonio Cuni
py3k-ify
Antonio Cuni
re-fix what was broken by 0000a290ac54 and 0954589ca523: on py3k, the repr() of longs is the same as their str(), i.e. the number without the L suffix
Antonio Cuni
partial backout of 83fc38544bca: we don't want rlib/ to diverge from default
Antonio Cuni
partial backout of f4af6649c12f: we don't want rlib/ to diverge from default
Antonio Cuni
fix another test broken by aa1d34b5a927
Antonio Cuni
fix this test which was broken by aa1d34b5a927
Antonio Cuni
py3k-ify
Antonio Cuni
these two tests don't make sense on py3k where we have only unicode strings
Antonio Cuni
don't use the L notation (not supported in py3k), and kill useless print
Antonio Cuni
don't use the L notation (not supported in py3k), and use a value which is big enough not to fit a word on 64bit cpus
Antonio Cuni
py3k-ify. The test still fail because we cannot have unicode exception messages yet, but apart from that it raises the correct exception
Antonio Cuni
py3k-ify
Antonio Cuni
kill this test, we no longer have longs
Antonio Cuni
slight differences between py2 and py3k wrt chars and unicodes
Antonio Cuni
this test does not make sense for py3k, we cannot use % on bytes
Antonio Cuni
add a comment explaining why this test fails
Ronny Pfannschmidt
implement applevel urandom for posix os
Antonio Cuni
py3k-ify one test, and skip a py2-only one
Antonio Cuni
py3k-ify
Antonio Cuni
py3k-ify
Antonio Cuni
fix typo and skip more tests
Ronny Pfannschmidt
whops, actually inline the 3.2 stdlib changes
Ronny Pfannschmidt
first clean merge from stdlib-unification
Antonio Cuni
merge heads
Antonio Cuni
we no longer have W_IntObject, but longs
Antonio Cuni
we no longer have W_IntObject, but longs
Antonio Cuni
don't try to support _continuation before having the core py3k working
Antonio Cuni
py3k_skip the failing zzpickle tests. Some are only needed for stackless, I'll fix them once the core py3k is working
Antonio Cuni
skip micronumpy tests, we are not interested in them right now
Antonio Cuni
skip one more test
Antonio Cuni
add an applevel py3k_skip, and py3k_skip some multidict tests
Antonio Cuni
add a py3k_skip helper, for tests which can be skipped now on py3k, but should be fixed sooner or later
Antonio Cuni
py3k-ify
Antonio Cuni
py3k-ify
Armin Rigo
Fix the optimization for sys.exc_info().
Antonio Cuni
kill this test, it does not make sense now that we have string vs bytes
Antonio Cuni
tweak the test to work in this branch
Antonio Cuni
the long type has gone
Antonio Cuni
fix syntax
Antonio Cuni
the usual wrap/wrapbytes confusion
Antonio Cuni
test and fix by disabling the unicodeliststragegy when converting from bytes
Antonio Cuni
kill this class altogheter: range() no longer returns a list, so most of these tests don't pass
Antonio Cuni
bah, kill leftover from the merge
Antonio Cuni
py3k fixes
Antonio Cuni
fix some of the results of the merge. StringListStrategy is called UnicodeListStrategy on py3k, and withrangelist has been removed (but we still keep the tests, just because)
Antonio Cuni
hg merge default
Antonio Cuni
kill unicode_w from W_AbstractStringObject: in py3k, byte strings are not convertible to unicode using the default econding
Antonio Cuni
skip these numpypy tests for now, the applevel part of numpy has to been ported to python3
Antonio Cuni
s/buffer/memoryview
Antonio Cuni
__nonzero__ has been renamed to __bool__ in py3k
Antonio Cuni
dtypes are now passed around as unicode
Antonio Cuni
add w_unicode to the FakeObjSpace, it's needed for translating itertools
Antonio Cuni
add some TODO list
Antonio Cuni
disable IntDictStrategy for now: we cannot use it, since now 'ints' are acutally 'longs'
Antonio Cuni
hg merge default
Antonio Cuni
enable the array and struct modules, for test_array_write. Else the test imports the corresponding modules in lib_pypy, which are now broken because we no longer have buffer (and I am not sure that fixing them is a good idea)
Antonio Cuni
kill this line, which was introduced by 794e8ba95264: we never want pathname to point to a pyc file, else update_code_filenames puts *.pyc inside co_filename. All the tests introduced by 794e8ba95264 seems to pass, so I don't know what was its original purpose
Antonio Cuni
this is not supposed to raise, but to work
Antonio Cuni
s/buffer/memoryview
Antonio Cuni
s/buffer/memoryview
Antonio Cuni
s/buffer/memoryview
Antonio Cuni
s/buffer/memoryview
Antonio Cuni
we no longer have buffer at applevel, kill this check
Antonio Cuni
s/buffer/memoryview
Antonio Cuni
typo
Antonio Cuni
s/buffer/memoryview
Antonio Cuni
we no longer have buffers, but we have memoryview instead
Antonio Cuni
we no longer have the applevel 'buffer', so no way the get the raw bytes of the unicode string. Hardcode the bytes because we know it's UCS-4, at least on linux
Antonio Cuni
on 32bits, the accum variable might overflow to a long; make sure to cast back so a signed int before passing it to _store_digits
Antonio Cuni
kill longs and u'' string
Antonio Cuni
our ints all inherits from W_AbstractIntObject now
Antonio Cuni
we no longer have oldstyle classes
Antonio Cuni
we cannot mix bytes and unicode when .join(); adapt one test and kill the other
Antonio Cuni
2to3
Antonio Cuni
bah, ints are now W_LongObject, so this multimethod was never called. Adapt it to deal with longs
Antonio Cuni
kill these two tests about longs: they are no longer there
Antonio Cuni
fix this test and add a comment
Antonio Cuni
kill longs, unicode, range() --> list(range())
Antonio Cuni
kill long and unicode
Antonio Cuni
kill longs
Antonio Cuni
format() now works only on str, kill the old __unicode__ stuff
Antonio Cuni
explicitly make a list
Antonio Cuni
adapt unicode-->str and str-->bytes
Antonio Cuni
we no longer have longs
Antonio Cuni
forbid comparisons between memoryviews. test_buffer now passes
Antonio Cuni
bah, of course when we do a slice we need to wrap() the resulting memoryview, not wrapbytes() it. This makes test_basic passing
Antonio Cuni
add a repr similar to cpython
Antonio Cuni
memoryview.__getitem__ must return bytes, not str
Antonio Cuni
kill the buffer() builtin; the internal implementation is still used by memoryview. Also, refactor the buffer tests to use memoryview when possible. test_buffer passes with -A
Antonio Cuni
skip rope marshal tests: ropes are broken in py3k
Antonio Cuni
use bytes, not str with marshal.loads
734 commits not shown.