darling-libobjc2/objc/slot.h
2013-01-31 17:40:53 +00:00

49 lines
1.8 KiB
C

#if defined(__clang__) && !defined(__OBJC_RUNTIME_INTERNAL__)
#pragma clang system_header
#endif
#ifndef __OBJC_SLOT_H_INCLUDED__
#define __OBJC_SLOT_H_INCLUDED__
/**
* The objc_slot structure is used to permit safe IMP caching. It is returned
* by the new lookup APIs. When you cache an IMP, you should store a copy of
* the version field and a pointer to the slot.
*
* The slot version is guaranteed never to be 0. When updating a cache, you
* should use code of the following form:
*
* 1) version = 0;
* 2) slot->cachedFor = receiver->isa;
* 3) slot_cache = slot;
* 4) version = slot->version;
*
* The runtime guarantees that the version in any cachable slot will never be
* 0. This should ensure that, if the version and cache pointer mismatch, the
* next access will cause a cache miss.
*
* When using a cached slot, you should compare the owner pointer to the isa
* pointer of the receiver and the message and the version of the slot to your
* cached version.
*/
struct objc_slot
{
/** The class to which this slot is attached (used internally). */
Class owner;
/** The class for which this slot was cached. Note that this can be
* modified by different cache owners, in different threads. Doing so may
* cause some cache misses, but if different methods are sending messages
* to the same object and sharing a cached slot then it may also improve
* cache hits. Profiling is probably required here. */
Class cachedFor;
/** The (typed) selector for the method identified by this slot. */
const char *types;
/** The current version. This changes if the method changes or if a
* subclass overrides this method, potentially invalidating this cache. */
int version;
/** The method pointer for this method. */
IMP method;
/** Selector for this method. */
SEL selector;
} OBJC_NONPORTABLE;
#endif // __OBJC_SLOT_H_INCLUDED__