Commits

Show all
Author Commit Message Labels Comments Date
Carl Friedrich Bolz
a pypy-c test for a counter on a class
Branches
new-dict-proxy
Carl Friedrich Bolz
make the remaining uses of w_type.dict_w go through the new interface
Branches
new-dict-proxy
Carl Friedrich Bolz
implement fast changing of classes: if an attribute of a class is overwritten (as opposed to defined for the first time) turn it into a cell object. writing to the cell won't invalidate caches.
Branches
new-dict-proxy
Lukas Diekmann
(cfbolz, l.diekmann): rewrite dict proxies as a new dict implementation. This is a lot simpler than the current hack. It changes behaviour vs. CPython slightly: the dictproxy looks and behaves (almost) like a normal dict.
Branches
new-dict-proxy
Lukas Diekmann
merge
Lukas Diekmann
(cfbolz, l.diekmann): fix dict.setdefault to only do one hash computation and dict lookup
Amaury Forgeot d'Arc
Fix the value according to the comment just below. (although 1Gb of L2 cache would certainly be nice)
Alex Gaynor
Try to fix wait3, how could it ever have worked?
Romain Guillebert
Adding os.wait3 to the standard library
Antonio Cuni
merge heads
Antonio Cuni
update the inlined pyrepl to revision 4d9968d3e7da
Armin Rigo
Move sys.platform out of RPython code.
Armin Rigo
from __future__ import with_statement
Armin Rigo
Pom pom pom. No tests :-((
Armin Rigo
merge heads
Armin Rigo
Oups. Fix.
Amaury Forgeot d'Arc
Let pypy import .pyw files on Windows
Amaury Forgeot d'Arc
Fix one test suite failure in test_import on Windows
Armin Rigo
merge heads
Armin Rigo
Simplify, clean-up and document this function a bit more.
Armin Rigo
hg merge jit-shadowstack (only one checkin so far).
Alex Gaynor
fix tests
Amaury Forgeot d'Arc
Add java implementation of str.rsplit()
Amaury Forgeot d'Arc
Implement str.rsplit for the cli backend
Amaury Forgeot d'Arc
Fixes
Amaury Forgeot d'Arc
Add RPython support for str.rsplit(char, maxsplit=-1)
Amaury Forgeot d'Arc
The 'maxsplit' argument of str.split(sep, maxsplit=-1) is now RPython.
Armin Rigo
Uh, we don't need to clear the shadow stack's memory (which is important in case we create a lot of threads). It is going to contain random stuff anyway after a few pushes and pops. No code should access it past the root_stack_top limit, and (at least nowadays) the shadow stack is not emitted by genc but allocated dynamically.
Branches
jit-shadowstack
Armin Rigo
A branch in which to play with supporting shadowstack in the JIT. A possible idea would be to keep the current structure in the JIT's backend, and put a marker in the shadowstack when we enter the assembler. It would let the assembler code run at full speed while getting rid of the asmgcc dependency.
Branches
jit-shadowstack
Antonio Cuni
make sure not to run _setup twice, else we might get infinite recursion between raw_input and _old_raw_input
  1. Prev
  2. Next