1. Pypy
  2. Untitled project
  3. pypy
  4. Branches

Branches

py3k-listview_str

View source
py3k-listview_str
  • Contributors
    1. Loading...
Author Commit Message Date Builds
250+ commits behind default.
Antonio Cuni
translation fix: we would need a BytesStrategy or a listview_unicode to make it working
Antonio Cuni
attempt to re-enable space.listview_str. It was disabled because some strategies implemented it for unicode and some for bytes, so translation failed. We decide that listview_str is only for bytes, and that we will need to grow the equivalent unicode ones
Antonio Cuni
a branch where to try to re-enable listview_str
Antonio Cuni
update to pyrepl c385792628a7 (py3k-readline branch); now even fancycompleter and colored tab-completion works :)
Philip Jenvey
add the surrogatepass error handler
Philip Jenvey
check for the magic encoding comment even w/ bytes and or PyCF_SOURCE_IS_UTF8 (which seems irrelevant now that the default encoding is utf8 anyway)
Philip Jenvey
clarify the intent of compile(), fix eval's bytes handling
Philip Jenvey
a couple more encoding related failures
Amaury Forgeot d'Arc
Update pyrepl from upstream (py3k-readline branch)
Amaury Forgeot d'Arc
io module: replace some ValueError by UnsupportedOperation.
Amaury Forgeot d'Arc
CPython issue 13343: Fix a crash when a lambda expression uses a global variable in the default value of a keyword-only argument: (lambda *, arg=GLOBAL_NAME: None)
Amaury Forgeot d'Arc
Skip an implementation detail, and cherry-pick change for CPython Issue #14177 that is not yet released and that PyPy already implements (marshal.loads rejects unicode strings)
Amaury Forgeot d'Arc
Also skip part of this test, like for 2.7
Amaury Forgeot d'Arc
Do the same changes as in the 2.7 test suite.
Amaury Forgeot d'Arc
test_itertools: skip some implementation details and accepts TypeError when the types don't match.
Philip Jenvey
kill refs to the exceptions module
Philip Jenvey
2to3 most of lib_pypy except _ctypes/numpypy/pyrepl
Philip Jenvey
update error message
Philip Jenvey
partly revert b59013f9587e, this part is needed for the recently fixed http://bugs.python.org/issue16013
Antonio Cuni
catch the correct exception type for both pypy and cpython
Antonio Cuni
we must ignore the encoding cookie even if we are compiling bytes
Antonio Cuni
merge heads
Antonio Cuni
finish the work started in 97e1b1af67ce: run list comprehensions in their own scope, as we do for genexp, setcomp and dictcomp
Antonio Cuni
in-progress: make list comprehensions in their own function; I just check this is to push it on my notebook :-)
Amaury Forgeot d'Arc
Fix translation
Amaury Forgeot d'Arc
Fix syntax error when importing hashlib, or random, or tempfile.
Amaury Forgeot d'Arc
Like CPython, avoid too large factorials.
Amaury Forgeot d'Arc
Fix tests in _multiprocessing
Amaury Forgeot d'Arc
Remove socket.fromfd(), and dup() is now a module-level function.
Amaury Forgeot d'Arc
remove _socket.makefile: it's now implemented in socket.py.
Amaury Forgeot d'Arc
Fix translation
Philip Jenvey
fix tests for CO_FUTURE_BARRY_AS_BDFL
Philip Jenvey
fill in the optional _string module's docs
Amaury Forgeot d'Arc
Implement {:a} format spec for str.format().
Amaury Forgeot d'Arc
Implement "Barry as BDFL" compile flag. It's not only a joke: this is how <> is now disallowed in normal Python3 code.
Amaury Forgeot d'Arc
In "raise exc from cause", cause can be a exception type, and must be normalized: call it to create an instance.
Amaury Forgeot d'Arc
Fix time module to pass tests.
Amaury Forgeot d'Arc
Fix tests when run with python3 and -A.
Amaury Forgeot d'Arc
Some fixes in the pure-python version of _sha512.
Amaury Forgeot d'Arc
One more collect() to fix the test.
Amaury Forgeot d'Arc
__unicode__ don't exist anymore. __bytes__ is now the space method that wekrefs have to forward.
Amaury Forgeot d'Arc
Add support for the __bytes__ special method.
Amaury Forgeot d'Arc
csv: fix tests when run with -A, then fix the implementation.
Amaury Forgeot d'Arc
Implement the %a format code.
Amaury Forgeot d'Arc
Fix two CPython specific tests
Amaury Forgeot d'Arc
Use 'surrogateescape' error handling in fsencode/fsdecode functions.
Amaury Forgeot d'Arc
Fix own tests in module/_codecs
Amaury Forgeot d'Arc
__eq__ should be called even when the object is compared with itself. This accounts for at least two failures, in test_weakref.py and test_decimal.py
Amaury Forgeot d'Arc
Add a bunch of gc.collect() to test_weakref. Copied from the 2.7 branch.
Amaury Forgeot d'Arc
Add timeout to RLock.acquire()
Amaury Forgeot d'Arc
PyPy extension: inspect.getargs() works on built-in functions. Port the diff from 2.7.
Amaury Forgeot d'Arc
Fix one test in test_unicode.py
Amaury Forgeot d'Arc
Fix for -A tests which define a "cls.w_xxx"
Amaury Forgeot d'Arc
Fix %r formatting when given a string with printable character: the result is not ascii...
Amaury Forgeot d'Arc
Fix tests to pass with -A
Amaury Forgeot d'Arc
Remove pyexpat.Parser.returns_unicode (always True)
Amaury Forgeot d'Arc
pyexpat: add mappings for error messages.
Amaury Forgeot d'Arc
Fix test: don't mix unicode and byte strings in regexp search/replace.
Amaury Forgeot d'Arc
Fix test_app_sre when run with -A
Amaury Forgeot d'Arc
Found a way to have space.appexec() work with the -A option.
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
Skip an implementation detail about unitialized module.__dict__ (same change as for 2.7)
Amaury Forgeot d'Arc
Context manager for mmap.
Philip Jenvey
utilize isfinite
Philip Jenvey
py3 round
Philip Jenvey
merge default
Antonio Cuni
revert 8617a5ed3187 and fix it by actually returning NotImplemented from complex.__lt__&co. Not sure this is the 100% right way to do it, but it seems to do the job
Antonio Cuni
implement isdisjoint for keys/items dict views
Antonio Cuni
merge heads
Antonio Cuni
if a complex number is created by just a negative imaginary part, the real part should be -0.0, not 0.0. See CPython issue 9011: http://bugs.python.org/issue9011
Antonio Cuni
the fact that complex.__lt__ & co. returns NotImplemented is an impl detail
Antonio Cuni
in pypy we simply don't have complex.__floordiv__, but TypeError is raised anyway when trying to // a complex number
Philip Jenvey
we still need to special case unwrapped defaults too, now w/ a test
Philip Jenvey
avoid OverflowErrors in specialised int tuples now that our w_int is long
Philip Jenvey
fix for py3 map
Antonio Cuni
add one more edge case to complex parsing
Antonio Cuni
allow special unicode decimal digits and spaces in complex ctor
Antonio Cuni
long is no longer there in py3k (sorry for the word play)
Antonio Cuni
hg merge default
Antonio Cuni
hg merge default
Philip Jenvey
use py3 __next__/builtins.next()
Philip Jenvey
fix None handling
Philip Jenvey
fdopen now resides in os
Philip Jenvey
py3 syntax
Philip Jenvey
fix for py3 map
Philip Jenvey
don't export SSL_CTX_clear_options unless we have it
Philip Jenvey
fix None default
Amaury Forgeot d'Arc
compile() and eval() accept unicode strings with a "-*- coding: -*-" cookie, it is simply ignored.
Amaury Forgeot d'Arc
py3k does not allow utf-8 surrogates.
Philip Jenvey
fix new None handling in exec
Philip Jenvey
rework py3k's bytes workaround directly to WrappedDefault
Philip Jenvey
merge default
Philip Jenvey
merge default
Philip Jenvey
o all classes are newstyle now and thus cachable o py3 exception fix for test_change_methods
Philip Jenvey
__getslice__ -> __getitem__
Philip Jenvey
move the older py2 range tests (adapted to py3) into test_functional, they were misplaced into test__functools instead of reduce's tests. restore those too
Philip Jenvey
thread -> _thread
Amaury Forgeot d'Arc
Fix range.__repr__, test passes._
Amaury Forgeot d'Arc
Fix tests for _cffi_backend
Philip Jenvey
fill in missing export for RPyThreadAcquireLockTimed
Philip Jenvey
move py3's isfinite into rcomplex with the rest of the impls
Philip Jenvey
merge default
Amaury Forgeot d'Arc
map(), filter(), zip() now return iterators. Move the code and tests from the itertool module. Most of the tests now pass with and without -A.
Amaury Forgeot d'Arc
Fix many tests for execution with -A.
Amaury Forgeot d'Arc
Fix test_chr to pass with -A (CPython 3.2)
Amaury Forgeot d'Arc
Fix getattr unit tests.
Philip Jenvey
switch to builtin next
Antonio Cuni
getsizeof is an impl detail
Antonio Cuni
bah, of course we need to call unicode_w in the unicode set strategy
Antonio Cuni
this is a cpython impl detail
Antonio Cuni
allow unicode spaces in int literals
Philip Jenvey
update ascii per runicode changes
Philip Jenvey
add mode to TextIOWrapper repr
Philip Jenvey
simplify
Philip Jenvey
ensure __repr/str__ results are unicode
Philip Jenvey
_utf8 is no longer immutable
Philip Jenvey
SSL_CTX_sess_* are macros
Philip Jenvey
test range's handling of __index__, __int__ is no longer supported
Philip Jenvey
py3 syntax/changes
Philip Jenvey
fix translation
Philip Jenvey
kill fake space.w_long
Amaury Forgeot d'Arc
Fix some test_ztranslation failures, functions with unwrap_spec(str) could not be analyzed.
Amaury Forgeot d'Arc
Fix exception type, a test relies on it.
Amaury Forgeot d'Arc
CPython defines OP_ALL differently, see CPython Issue #13885.
Amaury Forgeot d'Arc
Oops, fix translation of _ssl module.
Amaury Forgeot d'Arc
Improve error message when the buffer interface is not supported.
Amaury Forgeot d'Arc
Fix keyword argument in socket.getaddrinfo()
Amaury Forgeot d'Arc
Add a gc_collect() in test_weakref
Amaury Forgeot d'Arc
Somehow SSLContext.options was defined but not exported for applevel.
Amaury Forgeot d'Arc
Fix error message for int('invalid'). Yes, CPython has a test for it.
Amaury Forgeot d'Arc
Fix tuple.__repr__ containing non-ascii strings.
Amaury Forgeot d'Arc
Fix implementation details in test_list.
Amaury Forgeot d'Arc
The default implementation of __ne__ should now call not(x==y). This explains why running CPython test suite always displayed "Warning -- threading._dangling was modified".
Antonio Cuni
implement all the cmp ops between dictviews and {set,frozenset} in terms of all_implemented_in, as CPython does. This is needed to be able tocompare .items() on dictionaries whose value are unhashable
Antonio Cuni
a failing test
Antonio Cuni
fix != between dict.{keys,items} and {set,frozenset}
Antonio Cuni
don't override the default 'ne' multimethod
Antonio Cuni
kill some of the workarounds introduced in 03e3cf83880b about == and != between sets and ANY, as they are no longer needed. This fixes the comparison of sets and dictviews
Antonio Cuni
rpython fix: use unicode strings for error messages, because we need to use %s on unicode args
Antonio Cuni
rpython fix
Antonio Cuni
bytes->unicode conversion for all of _csv
Antonio Cuni
the _csv module now expects unicode to read, not strings
Antonio Cuni
try to fix what was broken after the merge because of the translation-cleanup branch
Antonio Cuni
hg merge default
Antonio Cuni
implement containment operators for dict views
Antonio Cuni
test&fix for the only set operator which is not symmetric
Antonio Cuni
make sure that set operations on dict views work with sequences in general and when the dictview is the right operand. Also, reduce a bit the duplicated code with some metaprogramming
Antonio Cuni
implement the - operator between dict views and sets
Antonio Cuni
remove debug print
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
Remove a test which does not make sense now that int==long. Also fix an old __nonzero__ method.
Amaury Forgeot d'Arc
Fix VA_COPY on 32bit platforms, this lets all cpyext run to the end.
Antonio Cuni
replace StringSetStrategy with UnicodeSetStrategy; earlier, we but bytes into sets but we got out strings
Antonio Cuni
we need to put the chek for the empty set inside the try/finally, else the id() is never removed from currently_in_repr
Antonio Cuni
port some of the changes there are in lib-python/2.7; some tests still failing
Antonio Cuni
there is no consistency between raising TypeError or AttributeError in these cases
Amaury Forgeot d'Arc
Oops, fix merge.
Amaury Forgeot d'Arc
Translation fix.
Amaury Forgeot d'Arc
Fixes after merge.
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
Also fix ropeunicodeobject.
Amaury Forgeot d'Arc
Issue1262: str.__repr__ now passes all printable characters. Patch by arielby.
Amaury Forgeot d'Arc
Now that unicode.encode('utf8') can fail, it's important to build the cached utf8 copy only when requested. Otherwise chr(0xd800) crashes the interpreter... This is difficult to test unfortunately, because the untranslated version does not fail.
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
Issue1262 in-progress: add unicode.isprintable(). One test stil fails, until we use Unicode version 6.0. Patch by arielby.
Amaury Forgeot d'Arc
Issue1262 in-progress: lone surrogates are invalid in utf-8. One exception: when encoding, and only for narrow unicode builds. Patch by arielby
Amaury Forgeot d'Arc
Issue1262 in-progress: add isprintable() to the unicode database. Patch by arielby.
Amaury Forgeot d'Arc
Remove duplicate implementation of RAISE_VARARGS.
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
cpyext: add support for the new buffer interface. Most of the tests pass now...
Amaury Forgeot d'Arc
cpyext: easy fixes in structmember.py
Amaury Forgeot d'Arc
cpyext: Adapt cpython issue14325, already done in the 2.x branch. This is a 3.3 feature, but pypy needs it to avoid the gc delay destruction of arguments.
Amaury Forgeot d'Arc
cpyext: simple fix for test_weakref.py
Amaury Forgeot d'Arc
cpyext: easy fixes in test_typeobject.py
Amaury Forgeot d'Arc
cpyext: fixes in pyerror.py
Amaury Forgeot d'Arc
cpyext: py3k-ify test_sysmodule.py
Amaury Forgeot d'Arc
cpyext: fix test_thread
Amaury Forgeot d'Arc
Terminal functions usually work with bytes strings.
Antonio Cuni
fix test_raise.test_with_reraise_2 by restoring the current exception after having called contetmanager's __exit__()
Antonio Cuni
a failing test and the corresponding fix. We want to make sure that the exception state is cleared when the blocks are unrolled because of a break
Antonio Cuni
convert the interp level TabError into an applevel one
Antonio Cuni
detect inconsisten use of tab and spaces and raise an inter-level TabError in that case. We still need to convert it to applevel
Antonio Cuni
one more gc_collect
Antonio Cuni
hg merge default
Antonio Cuni
don't clear the exception state between yields inside generators
Antonio Cuni
one more gc_collect()
Antonio Cuni
hg merge
Antonio Cuni
insert a gc_collect
Antonio Cuni
marshal.loads should not accept unicode. This is part of CPython's checkin 4966907d3661, which has not be included in stdlib-vendor yet
Antonio Cuni
UnicodeDecodeError accepts also bytearray
Antonio Cuni
we don't care about the exact exception type, as cpython is inconsistent
Amaury Forgeot d'Arc
Fix test in _multibytescodec module, and probably translation as well.
Amaury Forgeot d'Arc
Some py3k fixes in the _ffi module
Antonio Cuni
fix getrandbits() for the case where rbigint.SUPPORT_INT128 is enabled. Previosuly it always returned 0 but the test incorrectly passed because it returned a non-normalized rbigint(0, sign=1), while the correct form for 0 has sign==0.
Antonio Cuni
merge
Antonio Cuni
add some gc_collect() before dereferencing a weakref which is supposed to be dead
Amaury Forgeot d'Arc
Fixes in the ast module, after branch merge.
Amaury Forgeot d'Arc
Run interpreter/astcompiler/tools/asdl_py.py to regenerate ast.py
Amaury Forgeot d'Arc
hg merge default
Antonio Cuni
test&fix for converting bytes to int
Antonio Cuni
split unicode_encode_utf_8 in the same way as str_decode_utf_8
Amaury Forgeot d'Arc
hg merge default
Amaury Forgeot d'Arc
Adapt _cffi_backend to python3.
Amaury Forgeot d'Arc
The \u is no more an escape sequence in raw unicode strings: r'\u' is a string of two characters.
Amaury Forgeot d'Arc
"pytest -A" also checks the PYTHON3 environment variable to find a python3 interpreter.
Antonio Cuni
emit&implement DELETE_DEREF. This is equivalent to CPython's revision 25cc226dc9bd
Antonio Cuni
add a TODO item
Antonio Cuni
fix the flow objspace
Antonio Cuni
store both the unicode string and the utf8-encoded bytes for unicode objects, to avoid to do the encoding every time we call identifier_w to it. In the future, we might optimize W_UnicodeObject by storing *only* the utf-8 encoded bytes but for now it is easier to keep both. This change also fixes the methodcache tests, because they rely on the identity of the unwrapped string returned by str_w().
Antonio Cuni
py3k_skip cffi and micronumpy tests for now
Antonio Cuni
implement space.str() in terms of space.unicode_w. This fixes the two previously failing tests added in this commit
Antonio Cuni
split rlib.runicode.str_decode_utf_8 in two, else we had annotation problems when calling it to implement .decode('utf-8')
Antonio Cuni
comment out this failing assertion for now; we should look at it later
Antonio Cuni
hg merge default
Antonio Cuni
merge default
Antonio Cuni
hg merge default
Antonio Cuni
hg merge default
Antonio Cuni
hg merge default
Antonio Cuni
use identifier_w also here
Antonio Cuni
missing method on the dummy space
Antonio Cuni
use identifier_w to unwrap the keywords
Antonio Cuni
add a new space method to unwrap unicode identifiers into UTF-8 encoded strings
Antonio Cuni
hg backout f2376a5c4e21
Antonio Cuni
fix test_argument now that we use again bytes string and the exception messages are supposed to be in utf-8
Antonio Cuni
backout a5e20b0caee4: Signature() now takes byte strings again
Antonio Cuni
backout f860431bfa26, we no longer want unicode
Antonio Cuni
backout 6a591d00373a, we no longer want unicode
Antonio Cuni
backout 3b7e2e228239, we no longer want unicode
Antonio Cuni
backout bd3326f15584, we no longer want unicode
Antonio Cuni
completely change the strategy for unicode identifiers: instead of internally storing them as rpython unicode strings, we store them as utf-8 encoded byte strings, and decode them to unicode only when necessary (e.g., to present them to the user)
Antonio Cuni
add a new %8 formatting code, to indicate that the argument is a utf-8 encoded string and that it should be decoded when building the exception message
Antonio Cuni
improve the error message
Antonio Cuni
one more place where to pass unicode to Signature()
Antonio Cuni
fix Signature()/unicode also here
Antonio Cuni
one more place where to pass unicode to Signature()
Antonio Cuni
one more place where to pass unicode to Signature()
Antonio Cuni
merge heads
Antonio Cuni
enforce the parameters of Signature() to be unicode
Antonio Cuni
don't complain if we pass None to something which expects unicode or str
Antonio Cuni
missing import
Antonio Cuni
disable view_as_kwargs for now
Antonio Cuni
tentative rpython fixes
Antonio Cuni
enforce the list of keywords to be unicode
Antonio Cuni
hg merge default
Antonio Cuni
this has been checked in by mistake
Antonio Cuni
hg merge default: this merge has been painful because default contains the virtual-arguments branch, and the refactoring of argument.py conflicted with the introduction of kwonly args which had already been done in py3k. I hope I did not break anything
Antonio Cuni
add support for more complex types in enforceargs, like [int] or {str:int}; also move the imports as late as possible, to prevent circular imports
Antonio Cuni
we have unicode keywords now
Antonio Cuni
two more annotation-time checks
990 commits not shown.