- More fixes related to the return references patch
- eval_string() and call_user_function_ex() still don't work.
- The libzend tree is untested and might not be stabl yet.

- Implement return by reference:
- In function declaration instead of the return statement
- In the assignment phase
- Implement ability to turn off support for call-time pass by refe

- Implement return by reference:
- In function declaration instead of the return statement
- In the assignment phase
- Implement ability to turn off support for call-time pass by reference

show more ...

Fix warnings surfacing in maintainer-mode.

- Forgot to check for BP_VAR_IS in the fix made for Thies' string offset
problem.

- Add use support (behaves like require, but will not use the same file twice)
- Add import support (behaves like include, but requires parentheses; will not
use the same file twice; Curre

- Add use support (behaves like require, but will not use the same file twice)
- Add import support (behaves like include, but requires parentheses; will not
use the same file twice; Currently, it is not yet properly implemented, and
only behaves like include)

show more ...

- Fix problem Thies reported. We by mistake separated variables which were
being fetched for read only.

That's a more thorough fix...

Fix bug #2817 - assignments to string offsets could erronously modify unrelated strings

That slipped away

- Moved var_uninit() for return_value to the beginning of DO_FCALL.
We forgot to do it for overloaded methods

- Functions whose return values aren't used have them freed in DO_FCALL
and don't need a special ZEND_FREE opcode following them anymore

- If a function's return value is unused then don't create a ZEND_FREE
opcode but free it after the function call in zend_execute.

- Forgot this

- Add support for used_return_value passed to internal functions.

- Fix crash with string offset assignments.

- Add support for BYREF_FORCE_REST

- Fix for Thies' leak and Andrei's crash

- Fix isset() with string offsets.

- Preliminary submit of Thie's patch. Will fix the rest on Windows
as this was added on UNIX with patch. Changed IS_SAME -> IS_IDENTICAL

- object.ptr was made NULL in DO_FCALL but wasn't restored. Right now I
push it in DO_FCALL and at the end of do_fcall_common it always gets
popped. We might be able to optimize it out.

- No idea why this bug didn't exist before. But I'm too tired to think of it.
During a regular do_fcall we need to set object.ptr to NULL and, thus,
push it in the beginning and pop it in

- No idea why this bug didn't exist before. But I'm too tired to think of it.
During a regular do_fcall we need to set object.ptr to NULL and, thus,
push it in the beginning and pop it in the end.
I hope this fix more or less cuts it. I just want to sleep :)

show more ...

- Didn't lower refcount when doing an internal function call linked to a regular object.

- Clean up a bit. Separate before the locking so that we can use SEPARATE_ZVAL
macro.

- Shouldn't be needed

- God damn this sucked. I hopefully fixed the problems with classes although
we might need to clean stuff up a bit.