Take out the half-private g_private_init() stuff and replace it with a
G_PRIVATE_INIT macro that allows specifying a GDestroyNotify.
Expose the GPrivate structure in a public header.
Add a g_private_replace() to (sort of) match the functionality of
g_static_mutex_set().
Improve the documentation.
Deprecate g_private_new().
We lack SRWLock on Windows XP, so we use CRITICAL_SECTION to emulate it
there. SRWLock is non-recursive, but CRITICAL_SECTION will happily
allow itself to be acquired multiple times by the same thread.
We need to detect if our second acquire succeeded because of the
recursive nature of CRITICAL_SECTION. In the case of a _lock()
operation, we would normally have deadlocked, so abort. In the case of
a _trylock() operation, we need to ensure that FALSE is properly
returned.
Problem caught by Matthias Clasen and Chun-wei Fan.
https://bugzilla.gnome.org/show_bug.cgi?id=660096
Switch 'self' 'join' and 'create' from using the vtable to being called
via normal g_system_thread_* internal API (implemented in each of
gthread-{posix,win32}.c).
Again, we can put NULL in the vtable since these were never used from
gthread.h.
Thread priorities were already documented as not working on Solaris, and
they are meaningless on Linux unless the process separately requests
realtime scheduling (and even then, it appears only to work as root).
We can safely put a NULL into the vtable for set_priority since nothing
outside of gthread.c ever calls this (and that call is gone).
The "unknown reason" that the native thread implementation was broken is
because functions in kernel32.dll are (obviously) following Microsoft's
__stdcall ABI, not the GCC ABI.
Change our function pointers to be __stdcall pointers and change our
emulated implementation to match.
We remove the macros while at the same time switching all libglib users
from g_private_new() to g_private_init(). We deal with the strange
expectations of the libglib code that g_private_* should work before the
GPrivate has been initialised with a temporary shim.
- expose the structure types for GLib internal use only
- avoid infinite recursion hazards by ensuring that GPrivate never
calls back into any other part of GLib
- substantially rework the Windows implementation so that it never
holds locks, contains no arbitrary limits and doesn't waste
100*sizeof(void*) per thread
We have to keep the macro hacks for the time being since some code
inside libglib depends on it.
The original GMutex/GCond rework patch introduced some temporary code to
cope with GLib's old approach to thread initialisation. These are no
longer required.
Now that nothing inside of GLib is using g_cond_new(), we can implement
it using GSlice. Since the implementations for POSIX and Windows are
now the same, move it to gthread.c.
Now that nothing inside of GLib is using g_mutex_new, we can implement
it using GSlice. Since the implementations for POSIX and Windows are
now the same, move it to gthread.c.
Do a substantial rework of the GMutex and GCond APIs.
- remove all of the macro indirection hackery which is no longer needed
since we dropped support for switchable thread implementations
- expose the structure types and add G_MUTEX_INIT and G_COND_INIT
static initialiser macros
- add g_mutex_init() and g_mutex_clear() for use when embedding GMutex
into another structure type and do the same for GCond as well
- avoid infinite recursion hazards by ensuring that neither GCond or
GMutex ever calls back into any other part of GLib
- substantially rework the Windows implementation of GCond and GMutex
to use the SRWLock and CONDITION_VARIABLE APIs present on Windows
2008/Vista and later, emulating these APIs on XP
