1. Julien Phalip
  2. pypy

Commits

Show all
Author Commit Message Date Builds
Maciej Fijalkowski
a bit of shuffling before giving up
Maciej Fijalkowski
work in progress not sure where to go from here
Maciej Fijalkowski
an API I want from a custom tracer (that can be run directly too)
Maciej Fijalkowski
remove unused import
Antonio Cuni
numpy arrays should be weakreferenceable
Antonio Cuni
add a writable __pypy_data__ attribute to ndarray: it will be useful to implement the numpy C API in pure python
Antonio Cuni
add a way to construct a numpy array from an existing buffer. PyPy-only implementation detail
Alex Gaynor
Merged upstream.
Alex Gaynor
Reorganize where the dont_look_inside is to push it to a lower level. Doesn't affect anything yet because the JIT never looks in here.
Philip Jenvey
emit BOMs w/ utf16/32 empty strings. a py3k incremental encoding test (test_codecs's test_decoder_state) actually relies on this
Armin Rigo
Reintroduce a test similar to the previous one, and fix broken code.
Armin Rigo
Kill callback_hook. Replace it with a simpler and more explicit solution, which is to decorate explicitly the callbacks that you want the JIT to treat specially. Btw, another potential use of this would be to have the JIT see some function which is deep inside @dont_look_inside code, but which causes more app-level Python code to be invoked. This can be useful, to let the JIT see a bit more than just the app-level Python code (e.g. the wrapping of arguments).
Armin Rigo
Add jitting to C callbacks only on demand, not systematically. The RPython callback must have _callback_hook_="somestring". Might fix issue #1363, caused by the callback hook getting in the way of the shadowstack for the thread's bootstrapper.
Armin Rigo
An assert that fails (unless I messed up with gdb) on jit+shadowstack translations.
Armin Rigo
Pushing and popping roots around gc_thread_run() and gc_thread_die() doesn't make sense. Fixed by asserting we don't actually have to push any. Fix a test. PyPy translation was fixed by 509f13a420c6.
Armin Rigo
Don't use finally to call gc_thread_die: the exception has really nowhere to go.
Armin Rigo
Try to fix the test, but it still fails
Armin Rigo
Fix
Armin Rigo
Fix
Armin Rigo
Translation fix
Matti Picus
remove debug cruft
Matti Picus
backout half of previous uncareful commit
Matti Picus
windows does not like invalid signums
Maciej Fijalkowski
merge
Maciej Fijalkowski
fix enforceargs on some rffi helpers
Armin Rigo
jitdriver(reds='auto'): fix it to only include the red vars that are alive across the jitdriver. Test. Move the logic to support, too.
Armin Rigo
Add comments (from stm-thread-2).
Matti Picus
document merged branch
Matti Picus
switch ' and " for windows
Armin Rigo
Fix the remaining tests that use FINISH with more than one argument.
  1. Prev
  2. Next