Commits

Show all
Author Commit Message Labels Comments Date
Jeff Terrace
Moved unpacking down to the type level and fixed non-negative slice inference so all tests pass now
Branches
numpy-fromstring
Jeff Terrace
Add StringObject to FakeSpace and revert hacked isinstance_w workaround for w_str
Branches
numpy-fromstring
Jeff Terrace
Add full support for the rest of fromstring() functionality
Branches
numpy-fromstring
Jeff Terrace
Fixes for string special case to fix compiling tests
Branches
numpy-fromstring
Jeff Terrace
Treat strings as a special case in coerce to fix failing tests
Branches
numpy-fromstring
Jeff Terrace
Changed coerce methods to work with strings. Added tests for it.
Branches
numpy-fromstring
Jeff Terrace
Fixed typo in W_UnsignedIntegerBox, exposed and wrote tests for uint8, uint16, uint32
Branches
numpy-fromstring
Armin Rigo
Forbid SomeObject there.
Alex Gaynor
some of these generate getfield_gc_pures now
Maciej Fijalkowski
Fail slightly bettter than a SyntaxError
Armin Rigo
Skip these tests in py.test -A.
Armin Rigo
fixes for test_app_main.
Armin Rigo
Baaaaah. Don't rely on "is" being true for two independently wrapped strings. It fails on py.py, for example.
Armin Rigo
test and fix
Armin Rigo
Improve the hack 062e9d06c908: revert the changes done in the RegisterManager, and instead write a more involved but cleaner solution in the FrameManager. As a bonus you get tests too. This solution should work even when assembling bridges, not just loops.
Armin Rigo
Python 2.5 compat
Maciej Fijalkowski
add -g so debugger can be used
Armin Rigo
Write a hack to distinguish between "true built-in modules" and "pseudo-extension built-in modules". The latters are the ones that are extension modules in CPython. We list the formers explicitly. True built-in modules are treated like CPython treats built-in modules, i.e. they always shadow any xx.py. The pseudo-extension built-in modules are treated like CPython treats extension modules, and are only found in sys.path order for the fake entry '.../lib_pypy/__extensions__' which we put just before 'lib_pypy' and the 'lib-python' entries.
Christian Tismer
windows fix (mattip, stakkars)
Armin Rigo
For minimark, use by default 4MB for the nursery size, instead of half the number found by poking in /proc/cpuinfo, which may be a bit bogus. Also, nowadays, that rule seems wrong, in the sense that we need at least a few MBs to get good results, and increasing it doesn't change a lot.
Armin Rigo
bah.
Armin Rigo
A failing test: keepalive_until_there is not really strong enough.
Armin Rigo
Kill this __init__, which nowadays contains just a duplicate of setting to w_seq --- already done in the parent __init__.
Lukas Diekmann
merge
Lukas Diekmann
(l.diekmann, cfbolz): Be more careful about unrolling getitems. also make one less copy
Armin Rigo
Hack: record at least some partial information about which frame locations are freed, and if a hint is present, try to allocate the spilled boxes into these free locations. The hints are initialized based on looking ahead at the final JUMP operation. The goal is to reduce the number of frame-to-frame MOVs that must be emitted in the JUMP, by spilling boxes directly to the correct location, when possible.
Armin Rigo
Revert this, committed by accident. (The other accidental commit is actually good and a complete detail, so I'm not going to uncommit it just to commit it again...)
Armin Rigo
Add a failing test for fixedview() on a list of floats that expands to too much stuff for now.
Alex Gaynor
merged upstream
Alex Gaynor
make the dynamic interiorfield optimization work correctly with unsigned values
  1. Prev
  2. Next