Removes the deprecation warning when using the old-style syntax.

The return value for ::`->=() is no longer void.

The new-style argument checks were broken in several places.

The types now match the documentation.

handle_error() throwing errors at cleanup time now generates a
"Error in handle_error in master object!" with a meaningful
backtrace instead of a "No error recovery context.".

When indexing out a program constant, make it into a function
pointer also if the resulting program hasn't finished compiling.

Thanks to Per Cederqvist <cederp@opera.com> for the report and
test case.

Fixes [bug 7664].

Wait for the compiler to complete before complaining about
attempts to clone unfinished programs.

Thanks to Jeff Hungerford <hungerf3-roxen3@house.ofdoom.com>
for the report.

Potential fix for [Pike mailinglist 14495]/[LysLysKOM 21645192]

setters/getters are marked as variables with special run_time_type.
object_equal_p() did not handle that case and ended up calling
low_is_equal with type PIKE_T_GET_SET which would lead to a
fatal.

This change makes object_equal_p() ignore getters.

It is used to get the length, address and size shift of a memory/buffer
object storage.

The PROGRAM_CLEAR_STORAGE flag now sets the new object flag
OBJECT_CLEAR_ON_EXIT (analogous to STRING_CLEAR_ON_EXIT) at
clone time.

Make it easier to keep track of state in the visit_*() API
by tracking when we start and stop visiting things.

All the visit_*() functions and macros should now propagate the
argument extra.

All the visit_*() functions now have all arguments.

Adds a flag indicating that the storage held by objects of the class
should be cleared prior to being freed. This is mainly to avoid having
secrets around longer than necessary.

FIXME: Consider adding a corresponding flag for objects to mark eg
some bignums for clearing.

destroy() could leak/fatal if the thread received a signal
(eg due to cleanup_thread_state()). Delay any such signal
until destroy() is done.

Fixes fatal "Shouldn't get an exception in destruct()." on exit in
async_tls_close_test in pikes compiled --with-cleanup-on-exit.

Freeing of the destroy_called_mark_hash needs to be delayed until
after the last gc() call.

Fixes SEGV on exit if the gc on exit needs to destruct something
with an lfun::destroy().

This feature has not worked since at least y2k
(5b15bb75)

In fact, it never did work very well even befor that, and has always
much slower than not running unlocked. Especially on multi-cpu
systems.

Dynamic modules are now simulated with static modules when
not available. This includes having stub *.so files that
are loaded with load_module().

This fixes issues with eg loading Nettle without dynamic modules.

Make sure that parent_storage() also works from exit event
handlers. Fixes fatal "Shouldn't get an exception in destruct()."

Also propagates the PROGRAM_HAS_C_METHODS flag on inherit.

allocating many blocks and deallocation happens non linearly.

Now that PIKE_T_INT is zero, all globals are already cleared
when call_c_initializers() is called, so we can remove the
now redundant code.

This speeds up assignment of a lot of object variable types,
svalue_is_zero is (comparatively) expensive.

This is to simplify renumbering of the types.

Avoid keeping track of if destroy() has been called for objects that
lack destroy().

Thanks to Chris Angelico <rosuav@gmail.com> for the report.