diff options
| -rw-r--r-- | dbus/dbus-internals.c | 29 |
1 files changed, 29 insertions, 0 deletions
diff --git a/dbus/dbus-internals.c b/dbus/dbus-internals.c index 2c433677..3e578eb5 100644 --- a/dbus/dbus-internals.c +++ b/dbus/dbus-internals.c @@ -177,6 +177,35 @@ * Unlocks a global lock */ +/* The build system should have checked for DBUS_SIZEOF_VOID_P */ +_DBUS_STATIC_ASSERT (sizeof (void *) == DBUS_SIZEOF_VOID_P); + +/* dbus currently assumes that function pointers are essentially + * interchangeable with data pointers. There's nothing special about + * DBusShutdownFunction, it's just an arbitrary function pointer type. + * If this assertion fails on your platform, some porting will be required. */ +_DBUS_STATIC_ASSERT (sizeof (void *) == sizeof (DBusShutdownFunction)); +_DBUS_STATIC_ASSERT (_DBUS_ALIGNOF (void *) == _DBUS_ALIGNOF (DBusShutdownFunction)); + +/* This is meant to be true by definition. */ +_DBUS_STATIC_ASSERT (sizeof (void *) == sizeof (intptr_t)); +_DBUS_STATIC_ASSERT (sizeof (void *) == sizeof (uintptr_t)); + +/* + * Some frequent assumptions that we should *avoid* making include these, + * all of which are false on CHERI (which has 128-bit tagged pointers, + * but a 64-bit address space and therefore 64-bit sizes): + * + * sizeof (void *) <= sizeof (size_t) + * sizeof (void *) <= 8 + * _DBUS_ALIGNOF (void *) <= 8 + * + * We should also avoid making these assumptions, although we don't currently + * know a concrete example of platforms where they're false: + * + * sizeof (ptrdiff_t) == sizeof (size_t) + */ + /** * Fixed "out of memory" error message, just to avoid * making up a different string every time and wasting |