Branches

gc-del

View source
gc-del
Author Commit Message Labels Comments Date
250+ commits behind default.
Armin Rigo
Same as c9b26c923fc6 again.
Armin Rigo
Remove this line, which simply causes an exception to be printed
Armin Rigo
Not super happy with super_invoke_finalizer in mapdict. It feels like
Armin Rigo
Fix the case of objects with mapdict, which overrides a number of
Armin Rigo
Fix the test, but still failing
Armin Rigo
Failing test
Armin Rigo
hg merge default
Armin Rigo
Workaround for the config mechanism
Armin Rigo
hg merge default
Armin Rigo
Fix
Armin Rigo
Fix test
Armin Rigo
Fix
Armin Rigo
hg merge default
Armin Rigo
Fix for 6fed217ef39c
Armin Rigo
A failing test.
Armin Rigo
when not translated, be careful about multiple spaces for
Armin Rigo
Fixes
Armin Rigo
Remove this GC from the test
Armin Rigo
Reorganize this document, making it clear that all GCs apart from
Armin Rigo
Kill the config options for the removed GC, and fix test_sandbox.
Armin Rigo
More fixes
Armin Rigo
Fixes
Armin Rigo
Convert to use minimark.
Armin Rigo
Add missing operation
Armin Rigo
can't use lookup() here, must use _lookup(): see test_methodcache
Armin Rigo
Remove the "import x" from the same file x...
Armin Rigo
Use minimark here.
Armin Rigo
Fix: directly output the error message here, because we're not in finalizer_perform().
Armin Rigo
Fix test
Armin Rigo
Only call this if cards are enabled.
Armin Rigo
Fix
Armin Rigo
More fixes
Armin Rigo
Fix test_newgc.
Armin Rigo
Remove the semispace, generational, and hybrid GCs, after discussing
Armin Rigo
Constant-fold some more checks of GCFLAG_HAS_CARDS for the case
Armin Rigo
Fix the JIT.
Armin Rigo
Document the limitations of __del__ methods.
Armin Rigo
Fix.
Armin Rigo
Minimal support for register_finalizer() in the JIT
Armin Rigo
Fixes. Now an -O2 translates :-)
Armin Rigo
Fix the oracle module (blindly)
Armin Rigo
Fixes
Armin Rigo
Fix pyexpat.
Armin Rigo
Fix _multiprocessing.
Armin Rigo
Fix cpyext/pyobject.py. Add an operation to destructor.py.
Armin Rigo
Fix _winreg (untested).
Armin Rigo
Fix cpyext.
Armin Rigo
Bah, fix this logic
Armin Rigo
Fix
Armin Rigo
Port again the changes from default to enable both a destructor and a
Armin Rigo
Document this design decision.
Armin Rigo
Fix _io to call close() between bytecodes too. A hack for tests.
Armin Rigo
Fix _ssl.
Armin Rigo
Fix cppyy.
Armin Rigo
Fix _ffi.
Armin Rigo
Another __del__
Armin Rigo
Fix
Armin Rigo
More missing imports.
Armin Rigo
Some more missing imports
Armin Rigo
hg merge default
Armin Rigo
Fixes
Armin Rigo
haaaack to stop complaining at interpreter exit
Armin Rigo
Fix test
Armin Rigo
Fix _file.
Armin Rigo
Fix _io.
Armin Rigo
Fix
Armin Rigo
Fix generator.
Armin Rigo
hg merge default
Armin Rigo
Mapdict requires WeakrefLifeline to be a subclass of W_Root.
Armin Rigo
Fixes for module/gc.
Armin Rigo
Weakrefs, some small simplifications.
Armin Rigo
Test that "del A.__del__" works.
Armin Rigo
Fixes
Armin Rigo
Now attaching a __del__ to a class after its creation works.
Armin Rigo
Progress: __del__ works again on new-style classes.
Armin Rigo
Fix sockets
Armin Rigo
Adapt rmmap.
Armin Rigo
Trying to make sure we don't run finalizers after CPython started to
Armin Rigo
Generalize rgc.register_finalizer() into an API for any W_Root object.
Armin Rigo
Fix old-style classes for the new model of __del__
Armin Rigo
Added missing file
Armin Rigo
Transplant from "default" the fixes done to dotviewer.
Armin Rigo
Document the new destructors-and-finalizers world.
Armin Rigo
Progress
Armin Rigo
Support the new finalizer interface in Boehm too.
Armin Rigo
A (probable?) fix for the last remaining XXX about finalizers in the MiniMark GC
Armin Rigo
Move around the special-mode flag. Now both test_gc and test_newgc pass
Armin Rigo
Fix test.
Armin Rigo
Found out a simpler solution.
Armin Rigo
intermediate checkin, probably going to be partially reverted
Armin Rigo
Add a test that fails consistently, and fix.
Armin Rigo
Hack until test_newgc compiles.
Armin Rigo
Support raw array manipulations too
Armin Rigo
An extra test.
Armin Rigo
Fix test
Armin Rigo
test_gc and test_transformed_gc pass now (with MiniMark only!)
Armin Rigo
In-progress: start on framework.py.
Armin Rigo
Next test.
Armin Rigo
Various efforts to make this test pass --- but it was actually miswritten.
Armin Rigo
Relax the test, allowing now several finalizers to be called by the same
Armin Rigo
Copy the logic for young objects with finalizers, for the
Armin Rigo
Record surviving young objects with finalizers.
Armin Rigo
Fixes
Armin Rigo
Phew, finally I have (maybe) found out the right algorithm...
Armin Rigo
No-op refactoring.
Armin Rigo
Random progress
Armin Rigo
Lock
Armin Rigo
More in-progress.
Armin Rigo
in-progress
Armin Rigo
Hack hack hack to support register_finalizer in the llinterp.
Armin Rigo
PyPy compatibility of this untranslated hack
Armin Rigo
Test and fix for the case where there is also a __del__.
Armin Rigo
in-progress
Armin Rigo
Write an untranslated version of register_finalizer().
Armin Rigo
Wrote the algorithm, which is substantially simpler than the previous one.
Armin Rigo
Start writing a new finalizer-order.rst.
Armin Rigo
Document a possible memory leak.
Armin Rigo
Kill confusing comment. Adapt the title.
Armin Rigo
A branch in which the __del__() are called after only one major collection