more cleanups to object-memory-usage stuff -------------------- ChangeLog entries follow: -------------------- lisp/ChangeLog addition: 2010-03-19 Ben Wing <ben@xemacs.org> * diagnose.el (show-object-memory-usage-stats): Rewrite to take into account non-lisp-storage statistics returned by garbage-collect-1 and friends. src/ChangeLog addition: 2010-03-19 Ben Wing <ben@xemacs.org> * alloc.c: * alloc.c (struct): * alloc.c (tick_lrecord_stats): * alloc.c (gc_sweep_1): * alloc.c (finish_object_memory_usage_stats): * alloc.c (object_memory_usage_stats): * alloc.c (compute_memusage_stats_length): Call new memory-usage mechanism at sweep time to compute extra memory utilization for all objects. Add up the values element-by- element to get an aggregrate set of statistics, where each is the sum of the values of a single statistic across different objects of the same type. At end of sweep time, call finish_object_memory_usage_stats() to add up all the aggreggrate stats that are related to non-Lisp memory storage to compute a single value, and add it to the list of values returned by `garbage-collect' and `object-memory-usage-stats'. * buffer.c (compute_buffer_text_usage): Don't crash on buffers without text (killed buffers?) and don't double-count indirect buffers. * elhash.c: * elhash.c (hash_table_objects_create): * elhash.c (vars_of_elhash): * symsinit.h: Add memory-usage method to count the size of `hentries'. * emacs.c (main_1): Call new functions in elhash.c, frame.c at init. * frame.c: * frame.c (compute_frame_usage): * frame.c (frame_memory_usage): * frame.c (frame_objects_create): * symsinit.h: Add memory-usage method to count gutter display structures, subwindow exposures. * gc.c (gc_finish): * lisp.h: Declare finish_object_memory_usage_stats(), call it in gc_finish(). * lrecord.h (struct lrecord_implementation): * lrecord.h (INIT_MEMORY_USAGE_STATS): New value in implementation struct to track number of non-Lisp-memory statistics. Computed in alloc.c.

/* Synched up with: FSF 19.31. */

#include "usg5-4.h"

#define IRIX6

#ifdef LIBS_SYSTEM
#undef LIBS_SYSTEM
#endif

#ifdef LIB_STANDARD
#undef LIB_STANDARD
#endif

#ifdef SYSTEM_TYPE
#undef SYSTEM_TYPE
#endif
#define SYSTEM_TYPE "irix"

#ifdef SETUP_SLAVE_PTY
#undef SETUP_SLAVE_PTY
#endif

/* jpff@maths.bath.ac.uk reports `struct exception' is not defined
 * on this system, so inhibit use of matherr.  */
#define NO_MATHERR

/* Tell process_send_signal to use VSUSP instead of VSWTCH.  */
#define PREFER_VSUSP

/* use K&R C */
/* XEmacs change -- use ANSI, not K&R */
#ifndef __GNUC__
#define C_SWITCH_SYSTEM "-xansi"
#endif

/* jackr@engr.sgi.com says that you can't mix different kinds of
 * signal-handling functions under IRIX 5.3.  I'm going to assume
 * that that was the reason this got broken.  Now that the
 * signal routines are fixed up, maybe this will work. --ben */
/* Nope, it doesn't.  I've tried lots of things; it must be
 * genuinely broken. */
/* XEmacs addition: People on IRIX 5.2 and IRIX 5.3 systems have
 * reported that they can't break out of (while t) using C-g or C-G.
 * This does not occur on other systems, so let's assume that SIGIO
 * is broken on these systems. */
#define BROKEN_SIGIO

/* #### Questionable define. */
#define IRIX

/* By Tor Arntsen <tor@spacetec.no> for XEmacs.
 * With the following kludge the above LD_SWITCH_SYSTEM will still work just 
 * fine even with USE_GCC, and additional tweaking of config.h or ymakefile 
 * is avoided. */
#ifdef NOT_C_CODE
# ifdef USE_GCC
#  undef LINKER
#  undef LIB_GCC
#  define LINKER "ld"
#  define LIB_GCC "`gcc --print`"
#  endif
#endif