Add proccontrol/memcache.h

docs/proccontrol/developer/memcache.h.rst

@@ -1,5 +1,79 @@
 .. _`sec:memcache.h`:
 
 memcache.h
-==========
+##########
 
+.. warning::
+
+  DON'T USE THE MEMCACHE FOR ARBITRARY MEMORY OPERATIONS!
+
+  This class is meant to assist the SysV parser and the
+  thread_db parser--where we're wrapping another library
+  that is not async aware.
+
+  The semantics are such that this class expects that
+  reads and writes may be restarted, and will cache
+  and not redo operations it has already seen.  This
+  makes a difference when you do operations such as:
+
+  | write 'A' -> 0x1000
+  | write 'B' -> 0x1000
+  | write 'A' -> 0x1000
+
+  Under this class 0x1000 will contain 'B' after these three
+  operations.  The second 'A' write would be detected to be a
+  duplicated of the first and dropped.  We really do want these
+  semantics if we're restarting operations (we expect a write 'B'
+  will follow the second write 'A'), but this is inappropriate
+  for general purpose use.
+
+  Update - The memcache can now store registers.  Just what
+  every memcache needs.
+
+
+.. cpp:enum:: async_ret_t
+
+  .. cpp:enumerator:: aret_error = 0
+  .. cpp:enumerator:: aret_success
+  .. cpp:enumerator:: aret_async
+
+.. cpp:enum:: token_t
+
+  .. cpp:enumerator:: token_none = 0
+  .. cpp:enumerator:: token_getmsg
+  .. cpp:enumerator:: token_seteventreporting
+  .. cpp:enumerator:: token_setevent
+  .. cpp:enumerator:: token_init
+
+.. cpp:class:: memEntry
+
+  .. cpp:function:: memEntry()
+  .. cpp:function:: memEntry(token_t token)
+  .. cpp:function:: memEntry(Dyninst::Address remote, void *local, unsigned long size, bool is_read, memCache *cache)
+  .. cpp:function:: memEntry(const memEntry *me, char* b)
+  .. cpp:function:: memEntry& operator=(const memEntry&) = delete
+  .. cpp:function:: memEntry(const memEntry&) = delete
+  .. cpp:function:: ~memEntry()
+  .. cpp:function:: Dyninst::Address getAddress() const
+  .. cpp:function:: char *getBuffer() const
+  .. cpp:function:: unsigned long getSize() const
+  .. cpp:function:: bool isRead() const
+  .. cpp:function:: bool isWrite() const
+  .. cpp:function:: bool isToken() const
+  .. cpp:function:: bool operator==(const memEntry &b) const
+
+
+.. cpp:class:: memCache
+
+  .. cpp:function:: memCache(int_process *p)
+  .. cpp:function:: ~memCache()
+  .. cpp:function:: async_ret_t readMemory(void *dest, Dyninst::Address src, unsigned long size,  std::set<mem_response::ptr> &resps, int_thread *thrd = NULL)
+  .. cpp:function:: async_ret_t writeMemory(Dyninst::Address dest, void *src, unsigned long size,  std::set<result_response::ptr> &resps, int_thread *thrd = NULL)
+  .. cpp:function:: async_ret_t getRegisters(int_thread *thr, int_registerPool &pool)
+  .. cpp:function:: void startMemTrace(int &record)
+  .. cpp:function:: void clear()
+  .. cpp:function:: bool hasPendingAsync()
+  .. cpp:function:: void getPendingAsyncs(std::set<response::ptr> &resps)
+  .. cpp:function:: void setSyncHandling(bool b)
+  .. cpp:function:: void markToken(token_t tk)
+  .. cpp:function:: void condense()

proccontrol/src/memcache.h

@@ -53,32 +53,6 @@ typedef enum {
    token_init
 } token_t;
 
-/**
- * DON'T USE THE MEMCACHE FOR ARBITRARY MEMORY OPERATIONS!
- *
- * This class is meant to assist the SysV parser and the 
- * thread_db parser--where we're wrapping another library
- * that is not async aware.  
- * 
- * The semantics are such that this class expects that
- * reads and writes may be restarted, and will cache
- * and not redo operations it has already seen.  This
- * makes a difference when you do operations such as:
- * 
- * write 'A' -> 0x1000
- * write 'B' -> 0x1000
- * write 'A' -> 0x1000
- *
- * Under this class 0x1000 will contain 'B' after these three
- * operations.  The second 'A' write would be detected to be a
- * duplicated of the first and dropped.  We really do want these
- * semantics if we're restarting operations (we expect a write 'B'
- * will follow the second write 'A'), but this is inappropriate
- * for general purpose use.
- * 
- * Update - The memcache can now store registers.  Just what
- * every memcache needs.
- **/
 class memCache;
 class memEntry {
    friend struct memEntry_cmp;