/* Storage allocation and gc for XEmacs Lisp interpreter. Copyright (C) 1985-1998 Free Software Foundation, Inc. Copyright (C) 1995 Sun Microsystems, Inc. Copyright (C) 1995, 1996 Ben Wing. This file is part of XEmacs. XEmacs is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2, or (at your option) any later version. XEmacs is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with XEmacs; see the file COPYING. If not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */ /* Synched up with: FSF 19.28, Mule 2.0. Substantially different from FSF. */ /* Authorship: FSF: Original version; a long time ago. Mly: Significantly rewritten to use new 3-bit tags and nicely abstracted object definitions, for 19.8. JWZ: Improved code to keep track of purespace usage and issue nice purespace and GC stats. Ben Wing: Cleaned up frob-block lrecord code, added error-checking and various changes for Mule, for 19.12. Added bit vectors for 19.13. Added lcrecord lists for 19.14. slb: Lots of work on the purification and dump time code. Synched Doug Lea malloc support from Emacs 20.2. og: Killed the purespace. Portable dumper (moved to dumper.c) */ #include #include "lisp.h" #include "backtrace.h" #include "buffer.h" #include "bytecode.h" #include "chartab.h" #include "device.h" #include "elhash.h" #include "events.h" #include "extents.h" #include "frame.h" #include "glyphs.h" #include "opaque.h" #include "redisplay.h" #include "specifier.h" #include "sysfile.h" #include "sysdep.h" #include "window.h" #include "console-stream.h" #ifdef DOUG_LEA_MALLOC #include #endif #ifdef PDUMP #include "dumper.h" #endif EXFUN (Fgarbage_collect, 0); #if 0 /* this is _way_ too slow to be part of the standard debug options */ #if defined(DEBUG_XEMACS) && defined(MULE) #define VERIFY_STRING_CHARS_INTEGRITY #endif #endif /* Define this to use malloc/free with no freelist for all datatypes, the hope being that some debugging tools may help detect freed memory references */ #ifdef USE_DEBUG_MALLOC /* Taking the above comment at face value -slb */ #include #define ALLOC_NO_POOLS #endif #ifdef DEBUG_XEMACS static Fixnum debug_allocation; static Fixnum debug_allocation_backtrace_length; #endif /* Number of bytes of consing done since the last gc */ EMACS_INT consing_since_gc; #define INCREMENT_CONS_COUNTER_1(size) (consing_since_gc += (size)) #define debug_allocation_backtrace() \ do { \ if (debug_allocation_backtrace_length > 0) \ debug_short_backtrace (debug_allocation_backtrace_length); \ } while (0) #ifdef DEBUG_XEMACS #define INCREMENT_CONS_COUNTER(foosize, type) \ do { \ if (debug_allocation) \ { \ stderr_out ("allocating %s (size %ld)\n", type, (long)foosize); \ debug_allocation_backtrace (); \ } \ INCREMENT_CONS_COUNTER_1 (foosize); \ } while (0) #define NOSEEUM_INCREMENT_CONS_COUNTER(foosize, type) \ do { \ if (debug_allocation > 1) \ { \ stderr_out ("allocating noseeum %s (size %ld)\n", type, (long)foosize); \ debug_allocation_backtrace (); \ } \ INCREMENT_CONS_COUNTER_1 (foosize); \ } while (0) #else #define INCREMENT_CONS_COUNTER(size, type) INCREMENT_CONS_COUNTER_1 (size) #define NOSEEUM_INCREMENT_CONS_COUNTER(size, type) \ INCREMENT_CONS_COUNTER_1 (size) #endif #define DECREMENT_CONS_COUNTER(size) do { \ consing_since_gc -= (size); \ if (consing_since_gc < 0) \ consing_since_gc = 0; \ } while (0) /* Number of bytes of consing since gc before another gc should be done. */ EMACS_INT gc_cons_threshold; /* Nonzero during gc */ int gc_in_progress; /* Number of times GC has happened at this level or below. * Level 0 is most volatile, contrary to usual convention. * (Of course, there's only one level at present) */ EMACS_INT gc_generation_number[1]; /* This is just for use by the printer, to allow things to print uniquely */ static int lrecord_uid_counter; /* Nonzero when calling certain hooks or doing other things where a GC would be bad */ int gc_currently_forbidden; /* Hooks. */ Lisp_Object Vpre_gc_hook, Qpre_gc_hook; Lisp_Object Vpost_gc_hook, Qpost_gc_hook; /* "Garbage collecting" */ Lisp_Object Vgc_message; Lisp_Object Vgc_pointer_glyph; static const char gc_default_message[] = "Garbage collecting"; Lisp_Object Qgarbage_collecting; /* Non-zero means we're in the process of doing the dump */ int purify_flag; #ifdef ERROR_CHECK_TYPECHECK Error_behavior ERROR_ME, ERROR_ME_NOT, ERROR_ME_WARN; #endif int c_readonly (Lisp_Object obj) { return POINTER_TYPE_P (XTYPE (obj)) && C_READONLY (obj); } int lisp_readonly (Lisp_Object obj) { return POINTER_TYPE_P (XTYPE (obj)) && LISP_READONLY (obj); } /* Maximum amount of C stack to save when a GC happens. */ #ifndef MAX_SAVE_STACK #define MAX_SAVE_STACK 0 /* 16000 */ #endif /* Non-zero means ignore malloc warnings. Set during initialization. */ int ignore_malloc_warnings; static void *breathing_space; void release_breathing_space (void) { if (breathing_space) { void *tmp = breathing_space; breathing_space = 0; xfree (tmp); } } /* malloc calls this if it finds we are near exhausting storage */ void malloc_warning (const char *str) { if (ignore_malloc_warnings) return; warn_when_safe (Qmemory, Qcritical, "%s\n" "Killing some buffers may delay running out of memory.\n" "However, certainly by the time you receive the 95%% warning,\n" "you should clean up, kill this Emacs, and start a new one.", str); } /* Called if malloc returns zero */ DOESNT_RETURN memory_full (void) { /* Force a GC next time eval is called. It's better to loop garbage-collecting (we might reclaim enough to win) than to loop beeping and barfing "Memory exhausted" */ consing_since_gc = gc_cons_threshold + 1; release_breathing_space (); /* Flush some histories which might conceivably contain garbalogical inhibitors. */ if (!NILP (Fboundp (Qvalues))) Fset (Qvalues, Qnil); Vcommand_history = Qnil; error ("Memory exhausted"); } /* like malloc and realloc but check for no memory left, and block input. */ #undef xmalloc void * xmalloc (size_t size) { void *val = malloc (size); if (!val && (size != 0)) memory_full (); return val; } #undef xcalloc static void * xcalloc (size_t nelem, size_t elsize) { void *val = calloc (nelem, elsize); if (!val && (nelem != 0)) memory_full (); return val; } void * xmalloc_and_zero (size_t size) { return xcalloc (size, sizeof (char)); } #undef xrealloc void * xrealloc (void *block, size_t size) { /* We must call malloc explicitly when BLOCK is 0, since some reallocs don't do this. */ void *val = block ? realloc (block, size) : malloc (size); if (!val && (size != 0)) memory_full (); return val; } void #ifdef ERROR_CHECK_MALLOC xfree_1 (void *block) #else xfree (void *block) #endif { #ifdef ERROR_CHECK_MALLOC /* Unbelievably, calling free() on 0xDEADBEEF doesn't cause an error until much later on for many system mallocs, such as the one that comes with Solaris 2.3. FMH!! */ assert (block != (void *) 0xDEADBEEF); assert (block); #endif /* ERROR_CHECK_MALLOC */ free (block); } #ifdef ERROR_CHECK_GC #if SIZEOF_INT == 4 typedef unsigned int four_byte_t; #elif SIZEOF_LONG == 4 typedef unsigned long four_byte_t; #elif SIZEOF_SHORT == 4 typedef unsigned short four_byte_t; #else What kind of strange-ass system are we running on? #endif static void deadbeef_memory (void *ptr, size_t size) { four_byte_t *ptr4 = (four_byte_t *) ptr; size_t beefs = size >> 2; /* In practice, size will always be a multiple of four. */ while (beefs--) (*ptr4++) = 0xDEADBEEF; } #else /* !ERROR_CHECK_GC */ #define deadbeef_memory(ptr, size) #endif /* !ERROR_CHECK_GC */ #undef xstrdup char * xstrdup (const char *str) { int len = strlen (str) + 1; /* for stupid terminating 0 */ void *val = xmalloc (len); if (val == 0) return 0; return (char *) memcpy (val, str, len); } #ifdef NEED_STRDUP char * strdup (const char *s) { return xstrdup (s); } #endif /* NEED_STRDUP */ static void * allocate_lisp_storage (size_t size) { return xmalloc (size); } /* lcrecords are chained together through their "next" field. After doing the mark phase, GC will walk this linked list and free any lcrecord which hasn't been marked. */ static struct lcrecord_header *all_lcrecords; void * alloc_lcrecord (size_t size, const struct lrecord_implementation *implementation) { struct lcrecord_header *lcheader; type_checking_assert ((implementation->static_size == 0 ? implementation->size_in_bytes_method != NULL : implementation->static_size == size) && (! implementation->basic_p) && (! (implementation->hash == NULL && implementation->equal != NULL))); lcheader = (struct lcrecord_header *) allocate_lisp_storage (size); set_lheader_implementation (&lcheader->lheader, implementation); lcheader->next = all_lcrecords; #if 1 /* mly prefers to see small ID numbers */ lcheader->uid = lrecord_uid_counter++; #else /* jwz prefers to see real addrs */ lcheader->uid = (int) &lcheader; #endif lcheader->free = 0; all_lcrecords = lcheader; INCREMENT_CONS_COUNTER (size, implementation->name); return lcheader; } #if 0 /* Presently unused */ /* Very, very poor man's EGC? * This may be slow and thrash pages all over the place. * Only call it if you really feel you must (and if the * lrecord was fairly recently allocated). * Otherwise, just let the GC do its job -- that's what it's there for */ void free_lcrecord (struct lcrecord_header *lcrecord) { if (all_lcrecords == lcrecord) { all_lcrecords = lcrecord->next; } else { struct lrecord_header *header = all_lcrecords; for (;;) { struct lrecord_header *next = header->next; if (next == lcrecord) { header->next = lrecord->next; break; } else if (next == 0) ABORT (); else header = next; } } if (lrecord->implementation->finalizer) lrecord->implementation->finalizer (lrecord, 0); xfree (lrecord); return; } #endif /* Unused */ static void disksave_object_finalization_1 (void) { struct lcrecord_header *header; for (header = all_lcrecords; header; header = header->next) { if (LHEADER_IMPLEMENTATION (&header->lheader)->finalizer && !header->free) LHEADER_IMPLEMENTATION (&header->lheader)->finalizer (header, 1); } } /************************************************************************/ /* Debugger support */ /************************************************************************/ /* Give gdb/dbx enough information to decode Lisp Objects. We make sure certain symbols are always defined, so gdb doesn't complain about expressions in src/.gdbinit. See src/.gdbinit or src/.dbxrc to see how this is used. */ EMACS_UINT dbg_valmask = ((1UL << VALBITS) - 1) << GCBITS; EMACS_UINT dbg_typemask = (1UL << GCTYPEBITS) - 1; #ifdef USE_UNION_TYPE unsigned char dbg_USE_UNION_TYPE = 1; #else unsigned char dbg_USE_UNION_TYPE = 0; #endif unsigned char dbg_valbits = VALBITS; unsigned char dbg_gctypebits = GCTYPEBITS; /* On some systems, the above definitions will be optimized away by the compiler or linker unless they are referenced in some function. */ long dbg_inhibit_dbg_symbol_deletion (void); long dbg_inhibit_dbg_symbol_deletion (void) { return (dbg_valmask + dbg_typemask + dbg_USE_UNION_TYPE + dbg_valbits + dbg_gctypebits); } /* Macros turned into functions for ease of debugging. Debuggers don't know about macros! */ int dbg_eq (Lisp_Object obj1, Lisp_Object obj2); int dbg_eq (Lisp_Object obj1, Lisp_Object obj2) { return EQ (obj1, obj2); } /************************************************************************/ /* Fixed-size type macros */ /************************************************************************/ /* For fixed-size types that are commonly used, we malloc() large blocks of memory at a time and subdivide them into chunks of the correct size for an object of that type. This is more efficient than malloc()ing each object separately because we save on malloc() time and overhead due to the fewer number of malloc()ed blocks, and also because we don't need any extra pointers within each object to keep them threaded together for GC purposes. For less common (and frequently large-size) types, we use lcrecords, which are malloc()ed individually and chained together through a pointer in the lcrecord header. lcrecords do not need to be fixed-size (i.e. two objects of the same type need not have the same size; however, the size of a particular object cannot vary dynamically). It is also much easier to create a new lcrecord type because no additional code needs to be added to alloc.c. Finally, lcrecords may be more efficient when there are only a small number of them. The types that are stored in these large blocks (or "frob blocks") are cons, float, compiled-function, symbol, marker, extent, event, and string. Note that strings are special in that they are actually stored in two parts: a structure containing information about the string, and the actual data associated with the string. The former structure (a struct Lisp_String) is a fixed-size structure and is managed the same way as all the other such types. This structure contains a pointer to the actual string data, which is stored in structures of type struct string_chars_block. Each string_chars_block consists of a pointer to a struct Lisp_String, followed by the data for that string, followed by another pointer to a Lisp_String, followed by the data for that string, etc. At GC time, the data in these blocks is compacted by searching sequentially through all the blocks and compressing out any holes created by unmarked strings. Strings that are more than a certain size (bigger than the size of a string_chars_block, although something like half as big might make more sense) are malloc()ed separately and not stored in string_chars_blocks. Furthermore, no one string stretches across two string_chars_blocks. Vectors are each malloc()ed separately, similar to lcrecords. In the following discussion, we use conses, but it applies equally well to the other fixed-size types. We store cons cells inside of cons_blocks, allocating a new cons_block with malloc() whenever necessary. Cons cells reclaimed by GC are put on a free list to be reallocated before allocating any new cons cells from the latest cons_block. Each cons_block is just under 2^n - MALLOC_OVERHEAD bytes long, since malloc (at least the versions in malloc.c and gmalloc.c) really allocates in units of powers of two and uses 4 bytes for its own overhead. What GC actually does is to search through all the cons_blocks, from the most recently allocated to the oldest, and put all cons cells that are not marked (whether or not they're already free) on a cons_free_list. The cons_free_list is a stack, and so the cons cells in the oldest-allocated cons_block end up at the head of the stack and are the first to be reallocated. If any cons_block is entirely free, it is freed with free() and its cons cells removed from the cons_free_list. Because the cons_free_list ends up basically in memory order, we have a high locality of reference (assuming a reasonable turnover of allocating and freeing) and have a reasonable probability of entirely freeing up cons_blocks that have been more recently allocated. This stage is called the "sweep stage" of GC, and is executed after the "mark stage", which involves starting from all places that are known to point to in-use Lisp objects (e.g. the obarray, where are all symbols are stored; the current catches and condition-cases; the backtrace list of currently executing functions; the gcpro list; etc.) and recursively marking all objects that are accessible. At the beginning of the sweep stage, the conses in the cons blocks are in one of three states: in use and marked, in use but not marked, and not in use (already freed). Any conses that are marked have been marked in the mark stage just executed, because as part of the sweep stage we unmark any marked objects. The way we tell whether or not a cons cell is in use is through the LRECORD_FREE_P macro. This uses a special lrecord type `lrecord_type_free', which is never associated with any valid object. Conses on the free_cons_list are threaded through a pointer stored in the conses themselves. Because the cons is still in a cons_block and needs to remain marked as not in use for the next time that GC happens, we need room to store both the "free" indicator and the chaining pointer. So this pointer is stored after the lrecord header (actually where C places a pointer after the lrecord header; they are not necessarily contiguous). This implies that all fixed-size types must be big enough to contain at least one pointer. This is true for all current fixed-size types, with the possible exception of Lisp_Floats, for which we define the meat of the struct using a union of a pointer and a double to ensure adequate space for the free list chain pointer. Some types of objects need additional "finalization" done when an object is converted from in use to not in use; this is the purpose of the ADDITIONAL_FREE_type macro. For example, markers need to be removed from the chain of markers that is kept in each buffer. This is because markers in a buffer automatically disappear if the marker is no longer referenced anywhere (the same does not apply to extents, however). WARNING: Things are in an extremely bizarre state when the ADDITIONAL_FREE_type macros are called, so beware! When ERROR_CHECK_GC is defined, we do things differently so as to maximize our chances of catching places where there is insufficient GCPROing. The thing we want to avoid is having an object that we're using but didn't GCPRO get freed by GC and then reallocated while we're in the process of using it -- this will result in something seemingly unrelated getting trashed, and is extremely difficult to track down. If the object gets freed but not reallocated, we can usually catch this because we set most of the bytes of a freed object to 0xDEADBEEF. (The lisp object type is set to the invalid type `lrecord_type_free', however, and a pointer used to chain freed objects together is stored after the lrecord header; we play some tricks with this pointer to make it more bogus, so crashes are more likely to occur right away.) We want freed objects to stay free as long as possible, so instead of doing what we do above, we maintain the free objects in a first-in first-out queue. We also don't recompute the free list each GC, unlike above; this ensures that the queue ordering is preserved. [This means that we are likely to have worse locality of reference, and that we can never free a frob block once it's allocated. (Even if we know that all cells in it are free, there's no easy way to remove all those cells from the free list because the objects on the free list are unlikely to be in memory order.)] Furthermore, we never take objects off the free list unless there's a large number (usually 1000, but varies depending on type) of them already on the list. This way, we ensure that an object that gets freed will remain free for the next 1000 (or whatever) times that an object of that type is allocated. */ #ifndef MALLOC_OVERHEAD #ifdef GNU_MALLOC #define MALLOC_OVERHEAD 0 #elif defined (rcheck) #define MALLOC_OVERHEAD 20 #else #define MALLOC_OVERHEAD 8 #endif #endif /* MALLOC_OVERHEAD */ #if !defined(HAVE_MMAP) || defined(DOUG_LEA_MALLOC) /* If we released our reserve (due to running out of memory), and we have a fair amount free once again, try to set aside another reserve in case we run out once more. This is called when a relocatable block is freed in ralloc.c. */ void refill_memory_reserve (void); void refill_memory_reserve (void) { if (breathing_space == 0) breathing_space = (char *) malloc (4096 - MALLOC_OVERHEAD); } #endif #ifdef ALLOC_NO_POOLS # define TYPE_ALLOC_SIZE(type, structtype) 1 #else # define TYPE_ALLOC_SIZE(type, structtype) \ ((2048 - MALLOC_OVERHEAD - sizeof (struct type##_block *)) \ / sizeof (structtype)) #endif /* ALLOC_NO_POOLS */ #define DECLARE_FIXED_TYPE_ALLOC(type, structtype) \ \ struct type##_block \ { \ struct type##_block *prev; \ structtype block[TYPE_ALLOC_SIZE (type, structtype)]; \ }; \ \ static struct type##_block *current_##type##_block; \ static int current_##type##_block_index; \ \ static Lisp_Free *type##_free_list; \ static Lisp_Free *type##_free_list_tail; \ \ static void \ init_##type##_alloc (void) \ { \ current_##type##_block = 0; \ current_##type##_block_index = \ countof (current_##type##_block->block); \ type##_free_list = 0; \ type##_free_list_tail = 0; \ } \ \ static int gc_count_num_##type##_in_use; \ static int gc_count_num_##type##_freelist #define ALLOCATE_FIXED_TYPE_FROM_BLOCK(type, result) do { \ if (current_##type##_block_index \ == countof (current_##type##_block->block)) \ { \ struct type##_block *AFTFB_new = (struct type##_block *) \ allocate_lisp_storage (sizeof (struct type##_block)); \ AFTFB_new->prev = current_##type##_block; \ current_##type##_block = AFTFB_new; \ current_##type##_block_index = 0; \ } \ (result) = \ &(current_##type##_block->block[current_##type##_block_index++]); \ } while (0) /* Allocate an instance of a type that is stored in blocks. TYPE is the "name" of the type, STRUCTTYPE is the corresponding structure type. */ #ifdef ERROR_CHECK_GC /* Note: if you get crashes in this function, suspect incorrect calls to free_cons() and friends. This happened once because the cons cell was not GC-protected and was getting collected before free_cons() was called. */ #define ALLOCATE_FIXED_TYPE_1(type, structtype, result) do { \ if (gc_count_num_##type##_freelist > \ MINIMUM_ALLOWED_FIXED_TYPE_CELLS_##type) \ { \ result = (structtype *) type##_free_list; \ /* Before actually using the chain pointer, \ we complement all its bits; see FREE_FIXED_TYPE(). */ \ type##_free_list = (Lisp_Free *) \ (~ (EMACS_UINT) (type##_free_list->chain)); \ gc_count_num_##type##_freelist--; \ } \ else \ ALLOCATE_FIXED_TYPE_FROM_BLOCK (type, result); \ MARK_LRECORD_AS_NOT_FREE (result); \ } while (0) #else /* !ERROR_CHECK_GC */ #define ALLOCATE_FIXED_TYPE_1(type, structtype, result) do { \ if (type##_free_list) \ { \ result = (structtype *) type##_free_list; \ type##_free_list = type##_free_list->chain; \ } \ else \ ALLOCATE_FIXED_TYPE_FROM_BLOCK (type, result); \ MARK_LRECORD_AS_NOT_FREE (result); \ } while (0) #endif /* !ERROR_CHECK_GC */ #define ALLOCATE_FIXED_TYPE(type, structtype, result) \ do \ { \ ALLOCATE_FIXED_TYPE_1 (type, structtype, result); \ INCREMENT_CONS_COUNTER (sizeof (structtype), #type); \ } while (0) #define NOSEEUM_ALLOCATE_FIXED_TYPE(type, structtype, result) \ do \ { \ ALLOCATE_FIXED_TYPE_1 (type, structtype, result); \ NOSEEUM_INCREMENT_CONS_COUNTER (sizeof (structtype), #type); \ } while (0) /* Lisp_Free is the type to represent a free list member inside a frob block of any lisp object type. */ typedef struct Lisp_Free { struct lrecord_header lheader; struct Lisp_Free *chain; } Lisp_Free; #define LRECORD_FREE_P(ptr) \ ((ptr)->lheader.type == lrecord_type_free) #define MARK_LRECORD_AS_FREE(ptr) \ ((void) ((ptr)->lheader.type = lrecord_type_free)) #ifdef ERROR_CHECK_GC #define MARK_LRECORD_AS_NOT_FREE(ptr) \ ((void) ((ptr)->lheader.type = lrecord_type_undefined)) #else #define MARK_LRECORD_AS_NOT_FREE(ptr) DO_NOTHING #endif #ifdef ERROR_CHECK_GC #define PUT_FIXED_TYPE_ON_FREE_LIST(type, structtype, ptr) do { \ if (type##_free_list_tail) \ { \ /* When we store the chain pointer, we complement all \ its bits; this should significantly increase its \ bogosity in case someone tries to use the value, and \ should make us crash faster if someone overwrites the \ pointer because when it gets un-complemented in \ ALLOCATED_FIXED_TYPE(), the resulting pointer will be \ extremely bogus. */ \ type##_free_list_tail->chain = \ (Lisp_Free *) ~ (EMACS_UINT) (ptr); \ } \ else \ type##_free_list = (Lisp_Free *) (ptr); \ type##_free_list_tail = (Lisp_Free *) (ptr); \ } while (0) #else /* !ERROR_CHECK_GC */ #define PUT_FIXED_TYPE_ON_FREE_LIST(type, structtype, ptr) do { \ ((Lisp_Free *) (ptr))->chain = type##_free_list; \ type##_free_list = (Lisp_Free *) (ptr); \ } while (0) \ #endif /* !ERROR_CHECK_GC */ /* TYPE and STRUCTTYPE are the same as in ALLOCATE_FIXED_TYPE(). */ #define FREE_FIXED_TYPE(type, structtype, ptr) do { \ structtype *FFT_ptr = (ptr); \ ADDITIONAL_FREE_##type (FFT_ptr); \ deadbeef_memory (FFT_ptr, sizeof (structtype)); \ PUT_FIXED_TYPE_ON_FREE_LIST (type, structtype, FFT_ptr); \ MARK_LRECORD_AS_FREE (FFT_ptr); \ } while (0) /* Like FREE_FIXED_TYPE() but used when we are explicitly freeing a structure through free_cons(), free_marker(), etc. rather than through the normal process of sweeping. We attempt to undo the changes made to the allocation counters as a result of this structure being allocated. This is not completely necessary but helps keep things saner: e.g. this way, repeatedly allocating and freeing a cons will not result in the consing-since-gc counter advancing, which would cause a GC and somewhat defeat the purpose of explicitly freeing. */ #define FREE_FIXED_TYPE_WHEN_NOT_IN_GC(type, structtype, ptr) \ do { FREE_FIXED_TYPE (type, structtype, ptr); \ DECREMENT_CONS_COUNTER (sizeof (structtype)); \ gc_count_num_##type##_freelist++; \ } while (0) /************************************************************************/ /* Cons allocation */ /************************************************************************/ DECLARE_FIXED_TYPE_ALLOC (cons, Lisp_Cons); /* conses are used and freed so often that we set this really high */ /* #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_cons 20000 */ #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_cons 2000 static Lisp_Object mark_cons (Lisp_Object obj) { if (NILP (XCDR (obj))) return XCAR (obj); mark_object (XCAR (obj)); return XCDR (obj); } static int cons_equal (Lisp_Object ob1, Lisp_Object ob2, int depth) { depth++; while (internal_equal (XCAR (ob1), XCAR (ob2), depth)) { ob1 = XCDR (ob1); ob2 = XCDR (ob2); if (! CONSP (ob1) || ! CONSP (ob2)) return internal_equal (ob1, ob2, depth); } return 0; } static const struct lrecord_description cons_description[] = { { XD_LISP_OBJECT, offsetof (Lisp_Cons, car) }, { XD_LISP_OBJECT, offsetof (Lisp_Cons, cdr) }, { XD_END } }; DEFINE_BASIC_LRECORD_IMPLEMENTATION ("cons", cons, mark_cons, print_cons, 0, cons_equal, /* * No `hash' method needed. * internal_hash knows how to * handle conses. */ 0, cons_description, Lisp_Cons); DEFUN ("cons", Fcons, 2, 2, 0, /* Create a new cons, give it CAR and CDR as components, and return it. */ (car, cdr)) { /* This cannot GC. */ Lisp_Object val; Lisp_Cons *c; ALLOCATE_FIXED_TYPE (cons, Lisp_Cons, c); set_lheader_implementation (&c->lheader, &lrecord_cons); XSETCONS (val, c); c->car = car; c->cdr = cdr; return val; } /* This is identical to Fcons() but it used for conses that we're going to free later, and is useful when trying to track down "real" consing. */ Lisp_Object noseeum_cons (Lisp_Object car, Lisp_Object cdr) { Lisp_Object val; Lisp_Cons *c; NOSEEUM_ALLOCATE_FIXED_TYPE (cons, Lisp_Cons, c); set_lheader_implementation (&c->lheader, &lrecord_cons); XSETCONS (val, c); XCAR (val) = car; XCDR (val) = cdr; return val; } DEFUN ("list", Flist, 0, MANY, 0, /* Return a newly created list with specified arguments as elements. Any number of arguments, even zero arguments, are allowed. */ (int nargs, Lisp_Object *args)) { Lisp_Object val = Qnil; Lisp_Object *argp = args + nargs; while (argp > args) val = Fcons (*--argp, val); return val; } Lisp_Object list1 (Lisp_Object obj0) { /* This cannot GC. */ return Fcons (obj0, Qnil); } Lisp_Object list2 (Lisp_Object obj0, Lisp_Object obj1) { /* This cannot GC. */ return Fcons (obj0, Fcons (obj1, Qnil)); } Lisp_Object list3 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2) { /* This cannot GC. */ return Fcons (obj0, Fcons (obj1, Fcons (obj2, Qnil))); } Lisp_Object cons3 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2) { /* This cannot GC. */ return Fcons (obj0, Fcons (obj1, obj2)); } Lisp_Object acons (Lisp_Object key, Lisp_Object value, Lisp_Object alist) { return Fcons (Fcons (key, value), alist); } Lisp_Object list4 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2, Lisp_Object obj3) { /* This cannot GC. */ return Fcons (obj0, Fcons (obj1, Fcons (obj2, Fcons (obj3, Qnil)))); } Lisp_Object list5 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2, Lisp_Object obj3, Lisp_Object obj4) { /* This cannot GC. */ return Fcons (obj0, Fcons (obj1, Fcons (obj2, Fcons (obj3, Fcons (obj4, Qnil))))); } Lisp_Object list6 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2, Lisp_Object obj3, Lisp_Object obj4, Lisp_Object obj5) { /* This cannot GC. */ return Fcons (obj0, Fcons (obj1, Fcons (obj2, Fcons (obj3, Fcons (obj4, Fcons (obj5, Qnil)))))); } DEFUN ("make-list", Fmake_list, 2, 2, 0, /* Return a new list of length LENGTH, with each element being OBJECT. */ (length, object)) { CHECK_NATNUM (length); { Lisp_Object val = Qnil; size_t size = XINT (length); while (size--) val = Fcons (object, val); return val; } } /************************************************************************/ /* Float allocation */ /************************************************************************/ #ifdef LISP_FLOAT_TYPE DECLARE_FIXED_TYPE_ALLOC (float, Lisp_Float); #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_float 1000 Lisp_Object make_float (double float_value) { Lisp_Object val; Lisp_Float *f; ALLOCATE_FIXED_TYPE (float, Lisp_Float, f); /* Avoid dump-time `uninitialized memory read' purify warnings. */ if (sizeof (struct lrecord_header) + sizeof (double) != sizeof (*f)) xzero (*f); set_lheader_implementation (&f->lheader, &lrecord_float); float_data (f) = float_value; XSETFLOAT (val, f); return val; } #endif /* LISP_FLOAT_TYPE */ /************************************************************************/ /* Vector allocation */ /************************************************************************/ static Lisp_Object mark_vector (Lisp_Object obj) { Lisp_Vector *ptr = XVECTOR (obj); int len = vector_length (ptr); int i; for (i = 0; i < len - 1; i++) mark_object (ptr->contents[i]); return (len > 0) ? ptr->contents[len - 1] : Qnil; } static size_t size_vector (const void *lheader) { return FLEXIBLE_ARRAY_STRUCT_SIZEOF (Lisp_Vector, Lisp_Object, contents, ((Lisp_Vector *) lheader)->size); } static int vector_equal (Lisp_Object obj1, Lisp_Object obj2, int depth) { int len = XVECTOR_LENGTH (obj1); if (len != XVECTOR_LENGTH (obj2)) return 0; { Lisp_Object *ptr1 = XVECTOR_DATA (obj1); Lisp_Object *ptr2 = XVECTOR_DATA (obj2); while (len--) if (!internal_equal (*ptr1++, *ptr2++, depth + 1)) return 0; } return 1; } static hashcode_t vector_hash (Lisp_Object obj, int depth) { return HASH2 (XVECTOR_LENGTH (obj), internal_array_hash (XVECTOR_DATA (obj), XVECTOR_LENGTH (obj), depth + 1)); } static const struct lrecord_description vector_description[] = { { XD_LONG, offsetof (Lisp_Vector, size) }, { XD_LISP_OBJECT_ARRAY, offsetof (Lisp_Vector, contents), XD_INDIRECT(0, 0) }, { XD_END } }; DEFINE_LRECORD_SEQUENCE_IMPLEMENTATION("vector", vector, mark_vector, print_vector, 0, vector_equal, vector_hash, vector_description, size_vector, Lisp_Vector); /* #### should allocate `small' vectors from a frob-block */ static Lisp_Vector * make_vector_internal (size_t sizei) { /* no vector_next */ size_t sizem = FLEXIBLE_ARRAY_STRUCT_SIZEOF (Lisp_Vector, Lisp_Object, contents, sizei); Lisp_Vector *p = (Lisp_Vector *) alloc_lcrecord (sizem, &lrecord_vector); p->size = sizei; return p; } Lisp_Object make_vector (size_t length, Lisp_Object object) { Lisp_Vector *vecp = make_vector_internal (length); Lisp_Object *p = vector_data (vecp); while (length--) *p++ = object; { Lisp_Object vector; XSETVECTOR (vector, vecp); return vector; } } DEFUN ("make-vector", Fmake_vector, 2, 2, 0, /* Return a new vector of length LENGTH, with each element being OBJECT. See also the function `vector'. */ (length, object)) { CONCHECK_NATNUM (length); return make_vector (XINT (length), object); } DEFUN ("vector", Fvector, 0, MANY, 0, /* Return a newly created vector with specified arguments as elements. Any number of arguments, even zero arguments, are allowed. */ (int nargs, Lisp_Object *args)) { Lisp_Vector *vecp = make_vector_internal (nargs); Lisp_Object *p = vector_data (vecp); while (nargs--) *p++ = *args++; { Lisp_Object vector; XSETVECTOR (vector, vecp); return vector; } } Lisp_Object vector1 (Lisp_Object obj0) { return Fvector (1, &obj0); } Lisp_Object vector2 (Lisp_Object obj0, Lisp_Object obj1) { Lisp_Object args[2]; args[0] = obj0; args[1] = obj1; return Fvector (2, args); } Lisp_Object vector3 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2) { Lisp_Object args[3]; args[0] = obj0; args[1] = obj1; args[2] = obj2; return Fvector (3, args); } #if 0 /* currently unused */ Lisp_Object vector4 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2, Lisp_Object obj3) { Lisp_Object args[4]; args[0] = obj0; args[1] = obj1; args[2] = obj2; args[3] = obj3; return Fvector (4, args); } Lisp_Object vector5 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2, Lisp_Object obj3, Lisp_Object obj4) { Lisp_Object args[5]; args[0] = obj0; args[1] = obj1; args[2] = obj2; args[3] = obj3; args[4] = obj4; return Fvector (5, args); } Lisp_Object vector6 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2, Lisp_Object obj3, Lisp_Object obj4, Lisp_Object obj5) { Lisp_Object args[6]; args[0] = obj0; args[1] = obj1; args[2] = obj2; args[3] = obj3; args[4] = obj4; args[5] = obj5; return Fvector (6, args); } Lisp_Object vector7 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2, Lisp_Object obj3, Lisp_Object obj4, Lisp_Object obj5, Lisp_Object obj6) { Lisp_Object args[7]; args[0] = obj0; args[1] = obj1; args[2] = obj2; args[3] = obj3; args[4] = obj4; args[5] = obj5; args[6] = obj6; return Fvector (7, args); } Lisp_Object vector8 (Lisp_Object obj0, Lisp_Object obj1, Lisp_Object obj2, Lisp_Object obj3, Lisp_Object obj4, Lisp_Object obj5, Lisp_Object obj6, Lisp_Object obj7) { Lisp_Object args[8]; args[0] = obj0; args[1] = obj1; args[2] = obj2; args[3] = obj3; args[4] = obj4; args[5] = obj5; args[6] = obj6; args[7] = obj7; return Fvector (8, args); } #endif /* unused */ /************************************************************************/ /* Bit Vector allocation */ /************************************************************************/ static Lisp_Object all_bit_vectors; /* #### should allocate `small' bit vectors from a frob-block */ static Lisp_Bit_Vector * make_bit_vector_internal (size_t sizei) { size_t num_longs = BIT_VECTOR_LONG_STORAGE (sizei); size_t sizem = FLEXIBLE_ARRAY_STRUCT_SIZEOF (Lisp_Bit_Vector, unsigned long, bits, num_longs); Lisp_Bit_Vector *p = (Lisp_Bit_Vector *) allocate_lisp_storage (sizem); set_lheader_implementation (&p->lheader, &lrecord_bit_vector); INCREMENT_CONS_COUNTER (sizem, "bit-vector"); bit_vector_length (p) = sizei; bit_vector_next (p) = all_bit_vectors; /* make sure the extra bits in the last long are 0; the calling functions might not set them. */ p->bits[num_longs - 1] = 0; XSETBIT_VECTOR (all_bit_vectors, p); return p; } Lisp_Object make_bit_vector (size_t length, Lisp_Object bit) { Lisp_Bit_Vector *p = make_bit_vector_internal (length); size_t num_longs = BIT_VECTOR_LONG_STORAGE (length); CHECK_BIT (bit); if (ZEROP (bit)) memset (p->bits, 0, num_longs * sizeof (long)); else { size_t bits_in_last = length & (LONGBITS_POWER_OF_2 - 1); memset (p->bits, ~0, num_longs * sizeof (long)); /* But we have to make sure that the unused bits in the last long are 0, so that equal/hash is easy. */ if (bits_in_last) p->bits[num_longs - 1] &= (1 << bits_in_last) - 1; } { Lisp_Object bit_vector; XSETBIT_VECTOR (bit_vector, p); return bit_vector; } } Lisp_Object make_bit_vector_from_byte_vector (unsigned char *bytevec, size_t length) { size_t i; Lisp_Bit_Vector *p = make_bit_vector_internal (length); for (i = 0; i < length; i++) set_bit_vector_bit (p, i, bytevec[i]); { Lisp_Object bit_vector; XSETBIT_VECTOR (bit_vector, p); return bit_vector; } } DEFUN ("make-bit-vector", Fmake_bit_vector, 2, 2, 0, /* Return a new bit vector of length LENGTH. with each bit set to BIT. BIT must be one of the integers 0 or 1. See also the function `bit-vector'. */ (length, bit)) { CONCHECK_NATNUM (length); return make_bit_vector (XINT (length), bit); } DEFUN ("bit-vector", Fbit_vector, 0, MANY, 0, /* Return a newly created bit vector with specified arguments as elements. Any number of arguments, even zero arguments, are allowed. Each argument must be one of the integers 0 or 1. */ (int nargs, Lisp_Object *args)) { int i; Lisp_Bit_Vector *p = make_bit_vector_internal (nargs); for (i = 0; i < nargs; i++) { CHECK_BIT (args[i]); set_bit_vector_bit (p, i, !ZEROP (args[i])); } { Lisp_Object bit_vector; XSETBIT_VECTOR (bit_vector, p); return bit_vector; } } /************************************************************************/ /* Compiled-function allocation */ /************************************************************************/ DECLARE_FIXED_TYPE_ALLOC (compiled_function, Lisp_Compiled_Function); #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_compiled_function 1000 static Lisp_Object make_compiled_function (void) { Lisp_Compiled_Function *f; Lisp_Object fun; ALLOCATE_FIXED_TYPE (compiled_function, Lisp_Compiled_Function, f); set_lheader_implementation (&f->lheader, &lrecord_compiled_function); f->stack_depth = 0; f->specpdl_depth = 0; f->flags.documentationp = 0; f->flags.interactivep = 0; f->flags.domainp = 0; /* I18N3 */ f->instructions = Qzero; f->constants = Qzero; f->arglist = Qnil; f->doc_and_interactive = Qnil; #ifdef COMPILED_FUNCTION_ANNOTATION_HACK f->annotated = Qnil; #endif XSETCOMPILED_FUNCTION (fun, f); return fun; } DEFUN ("make-byte-code", Fmake_byte_code, 4, MANY, 0, /* Return a new compiled-function object. Usage: (arglist instructions constants stack-depth &optional doc-string interactive) Note that, unlike all other emacs-lisp functions, calling this with five arguments is NOT the same as calling it with six arguments, the last of which is nil. If the INTERACTIVE arg is specified as nil, then that means that this function was defined with `(interactive)'. If the arg is not specified, then that means the function is not interactive. This is terrible behavior which is retained for compatibility with old `.elc' files which expect these semantics. */ (int nargs, Lisp_Object *args)) { /* In a non-insane world this function would have this arglist... (arglist instructions constants stack_depth &optional doc_string interactive) */ Lisp_Object fun = make_compiled_function (); Lisp_Compiled_Function *f = XCOMPILED_FUNCTION (fun); Lisp_Object arglist = args[0]; Lisp_Object instructions = args[1]; Lisp_Object constants = args[2]; Lisp_Object stack_depth = args[3]; Lisp_Object doc_string = (nargs > 4) ? args[4] : Qnil; Lisp_Object interactive = (nargs > 5) ? args[5] : Qunbound; if (nargs < 4 || nargs > 6) return Fsignal (Qwrong_number_of_arguments, list2 (intern ("make-byte-code"), make_int (nargs))); /* Check for valid formal parameter list now, to allow us to use SPECBIND_FAST_UNSAFE() later in funcall_compiled_function(). */ { EXTERNAL_LIST_LOOP_3 (symbol, arglist, tail) { CHECK_SYMBOL (symbol); if (EQ (symbol, Qt) || EQ (symbol, Qnil) || SYMBOL_IS_KEYWORD (symbol)) signal_simple_error_2 ("Invalid constant symbol in formal parameter list", symbol, arglist); } } f->arglist = arglist; /* `instructions' is a string or a cons (string . int) for a lazy-loaded function. */ if (CONSP (instructions)) { CHECK_STRING (XCAR (instructions)); CHECK_INT (XCDR (instructions)); } else { CHECK_STRING (instructions); } f->instructions = instructions; if (!NILP (constants)) CHECK_VECTOR (constants); f->constants = constants; CHECK_NATNUM (stack_depth); f->stack_depth = (unsigned short) XINT (stack_depth); #ifdef COMPILED_FUNCTION_ANNOTATION_HACK if (!NILP (Vcurrent_compiled_function_annotation)) f->annotated = Fcopy (Vcurrent_compiled_function_annotation); else if (!NILP (Vload_file_name_internal_the_purecopy)) f->annotated = Vload_file_name_internal_the_purecopy; else if (!NILP (Vload_file_name_internal)) { struct gcpro gcpro1; GCPRO1 (fun); /* don't let fun get reaped */ Vload_file_name_internal_the_purecopy = Ffile_name_nondirectory (Vload_file_name_internal); f->annotated = Vload_file_name_internal_the_purecopy; UNGCPRO; } #endif /* COMPILED_FUNCTION_ANNOTATION_HACK */ /* doc_string may be nil, string, int, or a cons (string . int). interactive may be list or string (or unbound). */ f->doc_and_interactive = Qunbound; #ifdef I18N3 if ((f->flags.domainp = !NILP (Vfile_domain)) != 0) f->doc_and_interactive = Vfile_domain; #endif if ((f->flags.interactivep = !UNBOUNDP (interactive)) != 0) { f->doc_and_interactive = (UNBOUNDP (f->doc_and_interactive) ? interactive : Fcons (interactive, f->doc_and_interactive)); } if ((f->flags.documentationp = !NILP (doc_string)) != 0) { f->doc_and_interactive = (UNBOUNDP (f->doc_and_interactive) ? doc_string : Fcons (doc_string, f->doc_and_interactive)); } if (UNBOUNDP (f->doc_and_interactive)) f->doc_and_interactive = Qnil; return fun; } /************************************************************************/ /* Symbol allocation */ /************************************************************************/ DECLARE_FIXED_TYPE_ALLOC (symbol, Lisp_Symbol); #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_symbol 1000 DEFUN ("make-symbol", Fmake_symbol, 1, 1, 0, /* Return a newly allocated uninterned symbol whose name is NAME. Its value and function definition are void, and its property list is nil. */ (name)) { Lisp_Object val; Lisp_Symbol *p; CHECK_STRING (name); ALLOCATE_FIXED_TYPE (symbol, Lisp_Symbol, p); set_lheader_implementation (&p->lheader, &lrecord_symbol); p->name = XSTRING (name); p->plist = Qnil; p->value = Qunbound; p->function = Qunbound; symbol_next (p) = 0; XSETSYMBOL (val, p); return val; } /************************************************************************/ /* Extent allocation */ /************************************************************************/ DECLARE_FIXED_TYPE_ALLOC (extent, struct extent); #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_extent 1000 struct extent * allocate_extent (void) { struct extent *e; ALLOCATE_FIXED_TYPE (extent, struct extent, e); set_lheader_implementation (&e->lheader, &lrecord_extent); extent_object (e) = Qnil; set_extent_start (e, -1); set_extent_end (e, -1); e->plist = Qnil; xzero (e->flags); extent_face (e) = Qnil; e->flags.end_open = 1; /* default is for endpoints to behave like markers */ e->flags.detachable = 1; return e; } /************************************************************************/ /* Event allocation */ /************************************************************************/ DECLARE_FIXED_TYPE_ALLOC (event, Lisp_Event); #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_event 1000 Lisp_Object allocate_event (void) { Lisp_Object val; Lisp_Event *e; ALLOCATE_FIXED_TYPE (event, Lisp_Event, e); set_lheader_implementation (&e->lheader, &lrecord_event); XSETEVENT (val, e); return val; } /************************************************************************/ /* Marker allocation */ /************************************************************************/ DECLARE_FIXED_TYPE_ALLOC (marker, Lisp_Marker); #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_marker 1000 DEFUN ("make-marker", Fmake_marker, 0, 0, 0, /* Return a new marker which does not point at any place. */ ()) { Lisp_Object val; Lisp_Marker *p; ALLOCATE_FIXED_TYPE (marker, Lisp_Marker, p); set_lheader_implementation (&p->lheader, &lrecord_marker); p->buffer = 0; p->memind = 0; marker_next (p) = 0; marker_prev (p) = 0; p->insertion_type = 0; XSETMARKER (val, p); return val; } Lisp_Object noseeum_make_marker (void) { Lisp_Object val; Lisp_Marker *p; NOSEEUM_ALLOCATE_FIXED_TYPE (marker, Lisp_Marker, p); set_lheader_implementation (&p->lheader, &lrecord_marker); p->buffer = 0; p->memind = 0; marker_next (p) = 0; marker_prev (p) = 0; p->insertion_type = 0; XSETMARKER (val, p); return val; } /************************************************************************/ /* String allocation */ /************************************************************************/ /* The data for "short" strings generally resides inside of structs of type string_chars_block. The Lisp_String structure is allocated just like any other Lisp object (except for vectors), and these are freelisted when they get garbage collected. The data for short strings get compacted, but the data for large strings do not. Previously Lisp_String structures were relocated, but this caused a lot of bus-errors because the C code didn't include enough GCPRO's for strings (since EVERY REFERENCE to a short string needed to be GCPRO'd so that the reference would get relocated). This new method makes things somewhat bigger, but it is MUCH safer. */ DECLARE_FIXED_TYPE_ALLOC (string, Lisp_String); /* strings are used and freed quite often */ /* #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_string 10000 */ #define MINIMUM_ALLOWED_FIXED_TYPE_CELLS_string 1000 static Lisp_Object mark_string (Lisp_Object obj) { Lisp_String *ptr = XSTRING (obj); if (CONSP (ptr->plist) && EXTENT_INFOP (XCAR (ptr->plist))) flush_cached_extent_info (XCAR (ptr->plist)); return ptr->plist; } static int string_equal (Lisp_Object obj1, Lisp_Object obj2, int depth) { Bytecount len; return (((len = XSTRING_LENGTH (obj1)) == XSTRING_LENGTH (obj2)) && !memcmp (XSTRING_DATA (obj1), XSTRING_DATA (obj2), len)); } static const struct lrecord_description string_description[] = { { XD_BYTECOUNT, offsetof (Lisp_String, size) }, { XD_OPAQUE_DATA_PTR, offsetof (Lisp_String, data), XD_INDIRECT(0, 1) }, { XD_LISP_OBJECT, offsetof (Lisp_String, plist) }, { XD_END } }; /* We store the string's extent info as the first element of the string's property list; and the string's MODIFF as the first or second element of the string's property list (depending on whether the extent info is present), but only if the string has been modified. This is ugly but it reduces the memory allocated for the string in the vast majority of cases, where the string is never modified and has no extent info. #### This means you can't use an int as a key in a string's plist. */ static Lisp_Object * string_plist_ptr (Lisp_Object string) { Lisp_Object *ptr = &XSTRING (string)->plist; if (CONSP (*ptr) && EXTENT_INFOP (XCAR (*ptr))) ptr = &XCDR (*ptr); if (CONSP (*ptr) && INTP (XCAR (*ptr))) ptr = &XCDR (*ptr); return ptr; } static Lisp_Object string_getprop (Lisp_Object string, Lisp_Object property) { return external_plist_get (string_plist_ptr (string), property, 0, ERROR_ME); } static int string_putprop (Lisp_Object string, Lisp_Object property, Lisp_Object value) { external_plist_put (string_plist_ptr (string), property, value, 0, ERROR_ME); return 1; } static int string_remprop (Lisp_Object string, Lisp_Object property) { return external_remprop (string_plist_ptr (string), property, 0, ERROR_ME); } static Lisp_Object string_plist (Lisp_Object string) { return *string_plist_ptr (string); } /* No `finalize', or `hash' methods. internal_hash() already knows how to hash strings and finalization is done with the ADDITIONAL_FREE_string macro, which is the standard way to do finalization when using SWEEP_FIXED_TYPE_BLOCK(). */ DEFINE_BASIC_LRECORD_IMPLEMENTATION_WITH_PROPS ("string", string, mark_string, print_string, 0, string_equal, 0, string_description, string_getprop, string_putprop, string_remprop, string_plist, Lisp_String); /* String blocks contain this many useful bytes. */ #define STRING_CHARS_BLOCK_SIZE \ ((Bytecount) (8192 - MALLOC_OVERHEAD - \ ((2 * sizeof (struct string_chars_block *)) \ + sizeof (EMACS_INT)))) /* Block header for small strings. */ struct string_chars_block { EMACS_INT pos; struct string_chars_block *next; struct string_chars_block *prev; /* Contents of string_chars_block->string_chars are interleaved string_chars structures (see below) and the actual string data */ unsigned char string_chars[STRING_CHARS_BLOCK_SIZE]; }; static struct string_chars_block *first_string_chars_block; static struct string_chars_block *current_string_chars_block; /* If SIZE is the length of a string, this returns how many bytes * the string occupies in string_chars_block->string_chars * (including alignment padding). */ #define STRING_FULLSIZE(size) \ ALIGN_SIZE (((size) + 1 + sizeof (Lisp_String *)),\ ALIGNOF (Lisp_String *)) #define BIG_STRING_FULLSIZE_P(fullsize) ((fullsize) >= STRING_CHARS_BLOCK_SIZE) #define BIG_STRING_SIZE_P(size) (BIG_STRING_FULLSIZE_P (STRING_FULLSIZE(size))) #define STRING_CHARS_FREE_P(ptr) ((ptr)->string == NULL) #define MARK_STRING_CHARS_AS_FREE(ptr) ((void) ((ptr)->string = NULL)) struct string_chars { Lisp_String *string; unsigned char chars[1]; }; struct unused_string_chars { Lisp_String *string; EMACS_INT fullsize; }; static void init_string_chars_alloc (void) { first_string_chars_block = xnew (struct string_chars_block); first_string_chars_block->prev = 0; first_string_chars_block->next = 0; first_string_chars_block->pos = 0; current_string_chars_block = first_string_chars_block; } static struct string_chars * allocate_string_chars_struct (Lisp_String *string_it_goes_with, EMACS_INT fullsize) { struct string_chars *s_chars; if (fullsize <= (countof (current_string_chars_block->string_chars) - current_string_chars_block->pos)) { /* This string can fit in the current string chars block */ s_chars = (struct string_chars *) (current_string_chars_block->string_chars + current_string_chars_block->pos); current_string_chars_block->pos += fullsize; } else { /* Make a new current string chars block */ struct string_chars_block *new_scb = xnew (struct string_chars_block); current_string_chars_block->next = new_scb; new_scb->prev = current_string_chars_block; new_scb->next = 0; current_string_chars_block = new_scb; new_scb->pos = fullsize; s_chars = (struct string_chars *) current_string_chars_block->string_chars; } s_chars->string = string_it_goes_with; INCREMENT_CONS_COUNTER (fullsize, "string chars"); return s_chars; } Lisp_Object make_uninit_string (Bytecount length) { Lisp_String *s; EMACS_INT fullsize = STRING_FULLSIZE (length); Lisp_Object val; assert (length >= 0 && fullsize > 0); /* Allocate the string header */ ALLOCATE_FIXED_TYPE (string, Lisp_String, s); set_lheader_implementation (&s->lheader, &lrecord_string); set_string_data (s, BIG_STRING_FULLSIZE_P (fullsize) ? xnew_array (Bufbyte, length + 1) : allocate_string_chars_struct (s, fullsize)->chars); set_string_length (s, length); s->plist = Qnil; set_string_byte (s, length, 0); XSETSTRING (val, s); return val; } #ifdef VERIFY_STRING_CHARS_INTEGRITY static void verify_string_chars_integrity (void); #endif /* Resize the string S so that DELTA bytes can be inserted starting at POS. If DELTA < 0, it means deletion starting at POS. If POS < 0, resize the string but don't copy any characters. Use this if you're planning on completely overwriting the string. */ void resize_string (Lisp_String *s, Bytecount pos, Bytecount delta) { Bytecount oldfullsize, newfullsize; #ifdef VERIFY_STRING_CHARS_INTEGRITY verify_string_chars_integrity (); #endif #ifdef ERROR_CHECK_BUFPOS if (pos >= 0) { assert (pos <= string_length (s)); if (delta < 0) assert (pos + (-delta) <= string_length (s)); } else { if (delta < 0) assert ((-delta) <= string_length (s)); } #endif /* ERROR_CHECK_BUFPOS */ if (delta == 0) /* simplest case: no size change. */ return; if (pos >= 0 && delta < 0) /* If DELTA < 0, the functions below will delete the characters before POS. We want to delete characters *after* POS, however, so convert this to the appropriate form. */ pos += -delta; oldfullsize = STRING_FULLSIZE (string_length (s)); newfullsize = STRING_FULLSIZE (string_length (s) + delta); if (BIG_STRING_FULLSIZE_P (oldfullsize)) { if (BIG_STRING_FULLSIZE_P (newfullsize)) { /* Both strings are big. We can just realloc(). But careful! If the string is shrinking, we have to memmove() _before_ realloc(), and if growing, we have to memmove() _after_ realloc() - otherwise the access is illegal, and we might crash. */ Bytecount len = string_length (s) + 1 - pos; if (delta < 0 && pos >= 0) memmove (string_data (s) + pos + delta, string_data (s) + pos, len); set_string_data (s, (Bufbyte *) xrealloc (string_data (s), string_length (s) + delta + 1)); if (delta > 0 && pos >= 0) memmove (string_data (s) + pos + delta, string_data (s) + pos, len); } else /* String has been demoted from BIG_STRING. */ { Bufbyte *new_data = allocate_string_chars_struct (s, newfullsize)->chars; Bufbyte *old_data = string_data (s); if (pos >= 0) { memcpy (new_data, old_data, pos); memcpy (new_data + pos + delta, old_data + pos, string_length (s) + 1 - pos); } set_string_data (s, new_data); xfree (old_data); } } else /* old string is small */ { if (oldfullsize == newfullsize) { /* special case; size change but the necessary allocation size won't change (up or down; code somewhere depends on there not being any unused allocation space, modulo any alignment constraints). */ if (pos >= 0) { Bufbyte *addroff = pos + string_data (s); memmove (addroff + delta, addroff, /* +1 due to zero-termination. */ string_length (s) + 1 - pos); } } else { Bufbyte *old_data = string_data (s); Bufbyte *new_data = BIG_STRING_FULLSIZE_P (newfullsize) ? xnew_array (Bufbyte, string_length (s) + delta + 1) : allocate_string_chars_struct (s, newfullsize)->chars; if (pos >= 0) { memcpy (new_data, old_data, pos); memcpy (new_data + pos + delta, old_data + pos, string_length (s) + 1 - pos); } set_string_data (s, new_data); { /* We need to mark this chunk of the string_chars_block as unused so that compact_string_chars() doesn't freak. */ struct string_chars *old_s_chars = (struct string_chars *) ((char *) old_data - offsetof (struct string_chars, chars)); /* Sanity check to make sure we aren't hosed by strange alignment/padding. */ assert (old_s_chars->string == s); MARK_STRING_CHARS_AS_FREE (old_s_chars); ((struct unused_string_chars *) old_s_chars)->fullsize = oldfullsize; } } } set_string_length (s, string_length (s) + delta); /* If pos < 0, the string won't be zero-terminated. Terminate now just to make sure. */ string_data (s)[string_length (s)] = '\0'; if (pos >= 0) { Lisp_Object string; XSETSTRING (string, s); /* We also have to adjust all of the extent indices after the place we did the change. We say "pos - 1" because adjust_extents() is exclusive of the starting position passed to it. */ adjust_extents (string, pos - 1, string_length (s), delta); } #ifdef VERIFY_STRING_CHARS_INTEGRITY verify_string_chars_integrity (); #endif } #ifdef MULE void set_string_char (Lisp_String *s, Charcount i, Emchar c) { Bufbyte newstr[MAX_EMCHAR_LEN]; Bytecount bytoff = charcount_to_bytecount (string_data (s), i); Bytecount oldlen = charcount_to_bytecount (string_data (s) + bytoff, 1); Bytecount newlen = set_charptr_emchar (newstr, c); if (oldlen != newlen) resize_string (s, bytoff, newlen - oldlen); /* Remember, string_data (s) might have changed so we can't cache it. */ memcpy (string_data (s) + bytoff, newstr, newlen); } #endif /* MULE */ DEFUN ("make-string", Fmake_string, 2, 2, 0, /* Return a new string consisting of LENGTH copies of CHARACTER. LENGTH must be a non-negative integer. */ (length, character)) { CHECK_NATNUM (length); CHECK_CHAR_COERCE_INT (character); { Bufbyte init_str[MAX_EMCHAR_LEN]; int len = set_charptr_emchar (init_str, XCHAR (character)); Lisp_Object val = make_uninit_string (len * XINT (length)); if (len == 1) /* Optimize the single-byte case */ memset (XSTRING_DATA (val), XCHAR (character), XSTRING_LENGTH (val)); else { size_t i; Bufbyte *ptr = XSTRING_DATA (val); for (i = XINT (length); i; i--) { Bufbyte *init_ptr = init_str; switch (len) { case 4: *ptr++ = *init_ptr++; case 3: *ptr++ = *init_ptr++; case 2: *ptr++ = *init_ptr++; case 1: *ptr++ = *init_ptr++; } } } return val; } } DEFUN ("string", Fstring, 0, MANY, 0, /* Concatenate all the argument characters and make the result a string. */ (int nargs, Lisp_Object *args)) { Bufbyte *storage = alloca_array (Bufbyte, nargs * MAX_EMCHAR_LEN); Bufbyte *p = storage; for (; nargs; nargs--, args++) { Lisp_Object lisp_char = *args; CHECK_CHAR_COERCE_INT (lisp_char); p += set_charptr_emchar (p, XCHAR (lisp_char)); } return make_string (storage, p - storage); } /* Take some raw memory, which MUST already be in internal format, and package it up into a Lisp string. */ Lisp_Object make_string (const Bufbyte *contents, Bytecount length) { Lisp_Object val; /* Make sure we find out about bad make_string's when they happen */ #if defined (ERROR_CHECK_BUFPOS) && defined (MULE) bytecount_to_charcount (contents, length); /* Just for the assertions */ #endif val = make_uninit_string (length); memcpy (XSTRING_DATA (val), contents, length); return val; } /* Take some raw memory, encoded in some external data format, and convert it into a Lisp string. */ Lisp_Object make_ext_string (const Extbyte *contents, EMACS_INT length, Lisp_Object coding_system) { Lisp_Object string; TO_INTERNAL_FORMAT (DATA, (contents, length), LISP_STRING, string, coding_system); return string; } Lisp_Object build_string (const char *str) { /* Some strlen's crash and burn if passed null. */ return make_string ((const Bufbyte *) str, (str ? strlen(str) : 0)); } Lisp_Object build_ext_string (const char *str, Lisp_Object coding_system) { /* Some strlen's crash and burn if passed null. */ return make_ext_string ((const Extbyte *) str, (str ? strlen(str) : 0), coding_system); } Lisp_Object build_translated_string (const char *str) { return build_string (GETTEXT (str)); } Lisp_Object make_string_nocopy (const Bufbyte *contents, Bytecount length) { Lisp_String *s; Lisp_Object val; /* Make sure we find out about bad make_string_nocopy's when they happen */ #if defined (ERROR_CHECK_BUFPOS) && defined (MULE) bytecount_to_charcount (contents, length); /* Just for the assertions */ #endif /* Allocate the string header */ ALLOCATE_FIXED_TYPE (string, Lisp_String, s); set_lheader_implementation (&s->lheader, &lrecord_string); SET_C_READONLY_RECORD_HEADER (&s->lheader); s->plist = Qnil; set_string_data (s, (Bufbyte *)contents); set_string_length (s, length); XSETSTRING (val, s); return val; } /************************************************************************/ /* lcrecord lists */ /************************************************************************/ /* Lcrecord lists are used to manage the allocation of particular sorts of lcrecords, to avoid calling alloc_lcrecord() (and thus malloc() and garbage-collection junk) as much as possible. It is similar to the Blocktype class. It works like this: 1) Create an lcrecord-list object using make_lcrecord_list(). This is often done at initialization. Remember to staticpro_nodump this object! The arguments to make_lcrecord_list() are the same as would be passed to alloc_lcrecord(). 2) Instead of calling alloc_lcrecord(), call allocate_managed_lcrecord() and pass the lcrecord-list earlier created. 3) When done with the lcrecord, call free_managed_lcrecord(). The standard freeing caveats apply: ** make sure there are no pointers to the object anywhere! ** 4) Calling free_managed_lcrecord() is just like kissing the lcrecord goodbye as if it were garbage-collected. This means: -- the contents of the freed lcrecord are undefined, and the contents of something produced by allocate_managed_lcrecord() are undefined, just like for alloc_lcrecord(). -- the mark method for the lcrecord's type will *NEVER* be called on freed lcrecords. -- the finalize method for the lcrecord's type will be called at the time that free_managed_lcrecord() is called. */ static Lisp_Object mark_lcrecord_list (Lisp_Object obj) { struct lcrecord_list *list = XLCRECORD_LIST (obj); Lisp_Object chain = list->free; while (!NILP (chain)) { struct lrecord_header *lheader = XRECORD_LHEADER (chain); struct free_lcrecord_header *free_header = (struct free_lcrecord_header *) lheader; gc_checking_assert (/* There should be no other pointers to the free list. */ ! MARKED_RECORD_HEADER_P (lheader) && /* Only lcrecords should be here. */ ! LHEADER_IMPLEMENTATION (lheader)->basic_p && /* Only free lcrecords should be here. */ free_header->lcheader.free && /* The type of the lcrecord must be right. */ LHEADER_IMPLEMENTATION (lheader) == list->implementation && /* So must the size. */ (LHEADER_IMPLEMENTATION (lheader)->static_size == 0 || LHEADER_IMPLEMENTATION (lheader)->static_size == list->size) ); MARK_RECORD_HEADER (lheader); chain = free_header->chain; } return Qnil; } DEFINE_LRECORD_IMPLEMENTATION ("lcrecord-list", lcrecord_list, mark_lcrecord_list, internal_object_printer, 0, 0, 0, 0, struct lcrecord_list); Lisp_Object make_lcrecord_list (size_t size, const struct lrecord_implementation *implementation) { struct lcrecord_list *p = alloc_lcrecord_type (struct lcrecord_list, &lrecord_lcrecord_list); Lisp_Object val; p->implementation = implementation; p->size = size; p->free = Qnil; XSETLCRECORD_LIST (val, p); return val; } Lisp_Object allocate_managed_lcrecord (Lisp_Object lcrecord_list) { struct lcrecord_list *list = XLCRECORD_LIST (lcrecord_list); if (!NILP (list->free)) { Lisp_Object val = list->free; struct free_lcrecord_header *free_header = (struct free_lcrecord_header *) XPNTR (val); #ifdef ERROR_CHECK_GC struct lrecord_header *lheader = &free_header->lcheader.lheader; /* There should be no other pointers to the free list. */ assert (! MARKED_RECORD_HEADER_P (lheader)); /* Only lcrecords should be here. */ assert (! LHEADER_IMPLEMENTATION (lheader)->basic_p); /* Only free lcrecords should be here. */ assert (free_header->lcheader.free); /* The type of the lcrecord must be right. */ assert (LHEADER_IMPLEMENTATION (lheader) == list->implementation); /* So must the size. */ assert (LHEADER_IMPLEMENTATION (lheader)->static_size == 0 || LHEADER_IMPLEMENTATION (lheader)->static_size == list->size); #endif /* ERROR_CHECK_GC */ list->free = free_header->chain; free_header->lcheader.free = 0; return val; } else { Lisp_Object val; XSETOBJ (val, alloc_lcrecord (list->size, list->implementation)); return val; } } void free_managed_lcrecord (Lisp_Object lcrecord_list, Lisp_Object lcrecord) { struct lcrecord_list *list = XLCRECORD_LIST (lcrecord_list); struct free_lcrecord_header *free_header = (struct free_lcrecord_header *) XPNTR (lcrecord); struct lrecord_header *lheader = &free_header->lcheader.lheader; const struct lrecord_implementation *implementation = LHEADER_IMPLEMENTATION (lheader); /* Make sure the size is correct. This will catch, for example, putting a window configuration on the wrong free list. */ gc_checking_assert ((implementation->size_in_bytes_method ? implementation->size_in_bytes_method (lheader) : implementation->static_size) == list->size); if (implementation->finalizer) implementation->finalizer (lheader, 0); free_header->chain = list->free; free_header->lcheader.free = 1; list->free = lcrecord; } DEFUN ("purecopy", Fpurecopy, 1, 1, 0, /* Kept for compatibility, returns its argument. Old: Make a copy of OBJECT in pure storage. Recursively copies contents of vectors and cons cells. Does not copy symbols. */ (object)) { return object; } /************************************************************************/ /* Garbage Collection */ /************************************************************************/ /* All the built-in lisp object types are enumerated in `enum lrecord_type'. Additional ones may be defined by a module (none yet). We leave some room in `lrecord_implementations_table' for such new lisp object types. */ const struct lrecord_implementation *lrecord_implementations_table[(unsigned int)lrecord_type_last_built_in_type + MODULE_DEFINABLE_TYPE_COUNT]; unsigned int lrecord_type_count = (unsigned int)lrecord_type_last_built_in_type; /* Object marker functions are in the lrecord_implementation structure. But copying them to a parallel array is much more cache-friendly. This hack speeds up (garbage-collect) by about 5%. */ Lisp_Object (*lrecord_markers[countof (lrecord_implementations_table)]) (Lisp_Object); struct gcpro *gcprolist; /* We want the staticpros relocated, but not the pointers found therein. Hence we use a trivial description, as for pointerless objects. */ static const struct lrecord_description staticpro_description_1[] = { { XD_END } }; static const struct struct_description staticpro_description = { sizeof (Lisp_Object *), staticpro_description_1 }; static const struct lrecord_description staticpros_description_1[] = { XD_DYNARR_DESC (Lisp_Object_ptr_dynarr, &staticpro_description), { XD_END } }; static const struct struct_description staticpros_description = { sizeof (Lisp_Object_ptr_dynarr), staticpros_description_1 }; Lisp_Object_ptr_dynarr *staticpros; /* Mark the Lisp_Object at non-heap VARADDRESS as a root object for garbage collection, and for dumping. */ void staticpro (Lisp_Object *varaddress) { Dynarr_add (staticpros, varaddress); dump_add_root_object (varaddress); } Lisp_Object_ptr_dynarr *staticpros_nodump; /* Mark the Lisp_Object at non-heap VARADDRESS as a root object for garbage collection, but not for dumping. */ void staticpro_nodump (Lisp_Object *varaddress) { Dynarr_add (staticpros_nodump, varaddress); } #ifdef ERROR_CHECK_GC #define GC_CHECK_LHEADER_INVARIANTS(lheader) do { \ struct lrecord_header * GCLI_lh = (lheader); \ assert (GCLI_lh != 0); \ assert (GCLI_lh->type < lrecord_type_count); \ assert (! C_READONLY_RECORD_HEADER_P (GCLI_lh) || \ (MARKED_RECORD_HEADER_P (GCLI_lh) && \ LISP_READONLY_RECORD_HEADER_P (GCLI_lh))); \ } while (0) #else #define GC_CHECK_LHEADER_INVARIANTS(lheader) #endif /* Mark reference to a Lisp_Object. If the object referred to has not been seen yet, recursively mark all the references contained in it. */ void mark_object (Lisp_Object obj) { tail_recurse: /* Checks we used to perform */ /* if (EQ (obj, Qnull_pointer)) return; */ /* if (!POINTER_TYPE_P (XGCTYPE (obj))) return; */ /* if (PURIFIED (XPNTR (obj))) return; */ if (XTYPE (obj) == Lisp_Type_Record) { struct lrecord_header *lheader = XRECORD_LHEADER (obj); GC_CHECK_LHEADER_INVARIANTS (lheader); gc_checking_assert (LHEADER_IMPLEMENTATION (lheader)->basic_p || ! ((struct lcrecord_header *) lheader)->free); /* All c_readonly objects have their mark bit set, so that we only need to check the mark bit here. */ if (! MARKED_RECORD_HEADER_P (lheader)) { MARK_RECORD_HEADER (lheader); if (RECORD_MARKER (lheader)) { obj = RECORD_MARKER (lheader) (obj); if (!NILP (obj)) goto tail_recurse; } } } } /* mark all of the conses in a list and mark the final cdr; but DO NOT mark the cars. Use only for internal lists! There should never be other pointers to the cons cells, because if so, the cars will remain unmarked even when they maybe should be marked. */ void mark_conses_in_list (Lisp_Object obj) { Lisp_Object rest; for (rest = obj; CONSP (rest); rest = XCDR (rest)) { if (CONS_MARKED_P (XCONS (rest))) return; MARK_CONS (XCONS (rest)); } mark_object (rest); } /* Find all structures not marked, and free them. */ static int gc_count_num_bit_vector_used, gc_count_bit_vector_total_size; static int gc_count_bit_vector_storage; static int gc_count_num_short_string_in_use; static int gc_count_string_total_size; static int gc_count_short_string_total_size; /* static int gc_count_total_records_used, gc_count_records_total_size; */ /* stats on lcrecords in use - kinda kludgy */ static struct { int instances_in_use; int bytes_in_use; int instances_freed; int bytes_freed; int instances_on_free_list; } lcrecord_stats [countof (lrecord_implementations_table) + MODULE_DEFINABLE_TYPE_COUNT]; static void tick_lcrecord_stats (const struct lrecord_header *h, int free_p) { unsigned int type_index = h->type; if (((struct lcrecord_header *) h)->free) { gc_checking_assert (!free_p); lcrecord_stats[type_index].instances_on_free_list++; } else { const struct lrecord_implementation *implementation = LHEADER_IMPLEMENTATION (h); size_t sz = (implementation->size_in_bytes_method ? implementation->size_in_bytes_method (h) : implementation->static_size); if (free_p) { lcrecord_stats[type_index].instances_freed++; lcrecord_stats[type_index].bytes_freed += sz; } else { lcrecord_stats[type_index].instances_in_use++; lcrecord_stats[type_index].bytes_in_use += sz; } } } /* Free all unmarked records */ static void sweep_lcrecords_1 (struct lcrecord_header **prev, int *used) { struct lcrecord_header *header; int num_used = 0; /* int total_size = 0; */ xzero (lcrecord_stats); /* Reset all statistics to 0. */ /* First go through and call all the finalize methods. Then go through and free the objects. There used to be only one loop here, with the call to the finalizer occurring directly before the xfree() below. That is marginally faster but much less safe -- if the finalize method for an object needs to reference any other objects contained within it (and many do), we could easily be screwed by having already freed that other object. */ for (header = *prev; header; header = header->next) { struct lrecord_header *h = &(header->lheader); GC_CHECK_LHEADER_INVARIANTS (h); if (! MARKED_RECORD_HEADER_P (h) && ! header->free) { if (LHEADER_IMPLEMENTATION (h)->finalizer) LHEADER_IMPLEMENTATION (h)->finalizer (h, 0); } } for (header = *prev; header; ) { struct lrecord_header *h = &(header->lheader); if (MARKED_RECORD_HEADER_P (h)) { if (! C_READONLY_RECORD_HEADER_P (h)) UNMARK_RECORD_HEADER (h); num_used++; /* total_size += n->implementation->size_in_bytes (h);*/ /* #### May modify header->next on a C_READONLY lcrecord */ prev = &(header->next); header = *prev; tick_lcrecord_stats (h, 0); } else { struct lcrecord_header *next = header->next; *prev = next; tick_lcrecord_stats (h, 1); /* used to call finalizer right here. */ xfree (header); header = next; } } *used = num_used; /* *total = total_size; */ } static void sweep_bit_vectors_1 (Lisp_Object *prev, int *used, int *total, int *storage) { Lisp_Object bit_vector; int num_used = 0; int total_size = 0; int total_storage = 0; /* BIT_VECTORP fails because the objects are marked, which changes their implementation */ for (bit_vector = *prev; !EQ (bit_vector, Qzero); ) { Lisp_Bit_Vector *v = XBIT_VECTOR (bit_vector); int len = v->size; if (MARKED_RECORD_P (bit_vector)) { if (! C_READONLY_RECORD_HEADER_P(&(v->lheader))) UNMARK_RECORD_HEADER (&(v->lheader)); total_size += len; total_storage += MALLOC_OVERHEAD + FLEXIBLE_ARRAY_STRUCT_SIZEOF (Lisp_Bit_Vector, unsigned long, bits, BIT_VECTOR_LONG_STORAGE (len)); num_used++; /* #### May modify next on a C_READONLY bitvector */ prev = &(bit_vector_next (v)); bit_vector = *prev; } else { Lisp_Object next = bit_vector_next (v); *prev = next; xfree (v); bit_vector = next; } } *used = num_used; *total = total_size; *storage = total_storage; } /* And the Lord said: Thou shalt use the `c-backslash-region' command to make macros prettier. */ #ifdef ERROR_CHECK_GC #define SWEEP_FIXED_TYPE_BLOCK(typename, obj_type) \ do { \ struct typename##_block *SFTB_current; \ int SFTB_limit; \ int num_free = 0, num_used = 0; \ \ for (SFTB_current = current_##typename##_block, \ SFTB_limit = current_##typename##_block_index; \ SFTB_current; \ ) \ { \ int SFTB_iii; \ \ for (SFTB_iii = 0; SFTB_iii < SFTB_limit; SFTB_iii++) \ { \ obj_type *SFTB_victim = &(SFTB_current->block[SFTB_iii]); \ \ if (LRECORD_FREE_P (SFTB_victim)) \ { \ num_free++; \ } \ else if (C_READONLY_RECORD_HEADER_P (&SFTB_victim->lheader)) \ { \ num_used++; \ } \ else if (! MARKED_RECORD_HEADER_P (&SFTB_victim->lheader)) \ { \ num_free++; \ FREE_FIXED_TYPE (typename, obj_type, SFTB_victim); \ } \ else \ { \ num_used++; \ UNMARK_##typename (SFTB_victim); \ } \ } \ SFTB_current = SFTB_current->prev; \ SFTB_limit = countof (current_##typename##_block->block); \ } \ \ gc_count_num_##typename##_in_use = num_used; \ gc_count_num_##typename##_freelist = num_free; \ } while (0) #else /* !ERROR_CHECK_GC */ #define SWEEP_FIXED_TYPE_BLOCK(typename, obj_type) \ do { \ struct typename##_block *SFTB_current; \ struct typename##_block **SFTB_prev; \ int SFTB_limit; \ int num_free = 0, num_used = 0; \ \ typename##_free_list = 0; \ \ for (SFTB_prev = ¤t_##typename##_block, \ SFTB_current = current_##typename##_block, \ SFTB_limit = current_##typename##_block_index; \ SFTB_current; \ ) \ { \ int SFTB_iii; \ int SFTB_empty = 1; \ Lisp_Free *SFTB_old_free_list = typename##_free_list; \ \ for (SFTB_iii = 0; SFTB_iii < SFTB_limit; SFTB_iii++) \ { \ obj_type *SFTB_victim = &(SFTB_current->block[SFTB_iii]); \ \ if (LRECORD_FREE_P (SFTB_victim)) \ { \ num_free++; \ PUT_FIXED_TYPE_ON_FREE_LIST (typename, obj_type, SFTB_victim); \ } \ else if (C_READONLY_RECORD_HEADER_P (&SFTB_victim->lheader)) \ { \ SFTB_empty = 0; \ num_used++; \ } \ else if (! MARKED_RECORD_HEADER_P (&SFTB_victim->lheader)) \ { \ num_free++; \ FREE_FIXED_TYPE (typename, obj_type, SFTB_victim); \ } \ else \ { \ SFTB_empty = 0; \ num_used++; \ UNMARK_##typename (SFTB_victim); \ } \ } \ if (!SFTB_empty) \ { \ SFTB_prev = &(SFTB_current->prev); \ SFTB_current = SFTB_current->prev; \ } \ else if (SFTB_current == current_##typename##_block \ && !SFTB_current->prev) \ { \ /* No real point in freeing sole allocation block */ \ break; \ } \ else \ { \ struct typename##_block *SFTB_victim_block = SFTB_current; \ if (SFTB_victim_block == current_##typename##_block) \ current_##typename##_block_index \ = countof (current_##typename##_block->block); \ SFTB_current = SFTB_current->prev; \ { \ *SFTB_prev = SFTB_current; \ xfree (SFTB_victim_block); \ /* Restore free list to what it was before victim was swept */ \ typename##_free_list = SFTB_old_free_list; \ num_free -= SFTB_limit; \ } \ } \ SFTB_limit = countof (current_##typename##_block->block); \ } \ \ gc_count_num_##typename##_in_use = num_used; \ gc_count_num_##typename##_freelist = num_free; \ } while (0) #endif /* !ERROR_CHECK_GC */ static void sweep_conses (void) { #define UNMARK_cons(ptr) UNMARK_RECORD_HEADER (&((ptr)->lheader)) #define ADDITIONAL_FREE_cons(ptr) SWEEP_FIXED_TYPE_BLOCK (cons, Lisp_Cons); } /* Explicitly free a cons cell. */ void free_cons (Lisp_Cons *ptr) { #ifdef ERROR_CHECK_GC /* If the CAR is not an int, then it will be a pointer, which will always be four-byte aligned. If this cons cell has already been placed on the free list, however, its car will probably contain a chain pointer to the next cons on the list, which has cleverly had all its 0's and 1's inverted. This allows for a quick check to make sure we're not freeing something already freed. */ if (POINTER_TYPE_P (XTYPE (ptr->car))) ASSERT_VALID_POINTER (XPNTR (ptr->car)); #endif /* ERROR_CHECK_GC */ #ifndef ALLOC_NO_POOLS FREE_FIXED_TYPE_WHEN_NOT_IN_GC (cons, Lisp_Cons, ptr); #endif /* ALLOC_NO_POOLS */ } /* explicitly free a list. You **must make sure** that you have created all the cons cells that make up this list and that there are no pointers to any of these cons cells anywhere else. If there are, you will lose. */ void free_list (Lisp_Object list) { Lisp_Object rest, next; for (rest = list; !NILP (rest); rest = next) { next = XCDR (rest); free_cons (XCONS (rest)); } } /* explicitly free an alist. You **must make sure** that you have created all the cons cells that make up this alist and that there are no pointers to any of these cons cells anywhere else. If there are, you will lose. */ void free_alist (Lisp_Object alist) { Lisp_Object rest, next; for (rest = alist; !NILP (rest); rest = next) { next = XCDR (rest); free_cons (XCONS (XCAR (rest))); free_cons (XCONS (rest)); } } static void sweep_compiled_functions (void) { #define UNMARK_compiled_function(ptr) UNMARK_RECORD_HEADER (&((ptr)->lheader)) #define ADDITIONAL_FREE_compiled_function(ptr) SWEEP_FIXED_TYPE_BLOCK (compiled_function, Lisp_Compiled_Function); } #ifdef LISP_FLOAT_TYPE static void sweep_floats (void) { #define UNMARK_float(ptr) UNMARK_RECORD_HEADER (&((ptr)->lheader)) #define ADDITIONAL_FREE_float(ptr) SWEEP_FIXED_TYPE_BLOCK (float, Lisp_Float); } #endif /* LISP_FLOAT_TYPE */ static void sweep_symbols (void) { #define UNMARK_symbol(ptr) UNMARK_RECORD_HEADER (&((ptr)->lheader)) #define ADDITIONAL_FREE_symbol(ptr) SWEEP_FIXED_TYPE_BLOCK (symbol, Lisp_Symbol); } static void sweep_extents (void) { #define UNMARK_extent(ptr) UNMARK_RECORD_HEADER (&((ptr)->lheader)) #define ADDITIONAL_FREE_extent(ptr) SWEEP_FIXED_TYPE_BLOCK (extent, struct extent); } static void sweep_events (void) { #define UNMARK_event(ptr) UNMARK_RECORD_HEADER (&((ptr)->lheader)) #define ADDITIONAL_FREE_event(ptr) SWEEP_FIXED_TYPE_BLOCK (event, Lisp_Event); } static void sweep_markers (void) { #define UNMARK_marker(ptr) UNMARK_RECORD_HEADER (&((ptr)->lheader)) #define ADDITIONAL_FREE_marker(ptr) \ do { Lisp_Object tem; \ XSETMARKER (tem, ptr); \ unchain_marker (tem); \ } while (0) SWEEP_FIXED_TYPE_BLOCK (marker, Lisp_Marker); } /* Explicitly free a marker. */ void free_marker (Lisp_Marker *ptr) { /* Perhaps this will catch freeing an already-freed marker. */ gc_checking_assert (ptr->lheader.type == lrecord_type_marker); #ifndef ALLOC_NO_POOLS FREE_FIXED_TYPE_WHEN_NOT_IN_GC (marker, Lisp_Marker, ptr); #endif /* ALLOC_NO_POOLS */ } #if defined (MULE) && defined (VERIFY_STRING_CHARS_INTEGRITY) static void verify_string_chars_integrity (void) { struct string_chars_block *sb; /* Scan each existing string block sequentially, string by string. */ for (sb = first_string_chars_block; sb; sb = sb->next) { int pos = 0; /* POS is the index of the next string in the block. */ while (pos < sb->pos) { struct string_chars *s_chars = (struct string_chars *) &(sb->string_chars[pos]); Lisp_String *string; int size; int fullsize; /* If the string_chars struct is marked as free (i.e. the STRING pointer is NULL) then this is an unused chunk of string storage. (See below.) */ if (STRING_CHARS_FREE_P (s_chars)) { fullsize = ((struct unused_string_chars *) s_chars)->fullsize; pos += fullsize; continue; } string = s_chars->string; /* Must be 32-bit aligned. */ assert ((((int) string) & 3) == 0); size = string_length (string); fullsize = STRING_FULLSIZE (size); assert (!BIG_STRING_FULLSIZE_P (fullsize)); assert (string_data (string) == s_chars->chars); pos += fullsize; } assert (pos == sb->pos); } } #endif /* MULE && ERROR_CHECK_GC */ /* Compactify string chars, relocating the reference to each -- free any empty string_chars_block we see. */ static void compact_string_chars (void) { struct string_chars_block *to_sb = first_string_chars_block; int to_pos = 0; struct string_chars_block *from_sb; /* Scan each existing string block sequentially, string by string. */ for (from_sb = first_string_chars_block; from_sb; from_sb = from_sb->next) { int from_pos = 0; /* FROM_POS is the index of the next string in the block. */ while (from_pos < from_sb->pos) { struct string_chars *from_s_chars = (struct string_chars *) &(from_sb->string_chars[from_pos]); struct string_chars *to_s_chars; Lisp_String *string; int size; int fullsize; /* If the string_chars struct is marked as free (i.e. the STRING pointer is NULL) then this is an unused chunk of string storage. This happens under Mule when a string's size changes in such a way that its fullsize changes. (Strings can change size because a different-length character can be substituted for another character.) In this case, after the bogus string pointer is the "fullsize" of this entry, i.e. how many bytes to skip. */ if (STRING_CHARS_FREE_P (from_s_chars)) { fullsize = ((struct unused_string_chars *) from_s_chars)->fullsize; from_pos += fullsize; continue; } string = from_s_chars->string; assert (!(LRECORD_FREE_P (string))); size = string_length (string); fullsize = STRING_FULLSIZE (size); gc_checking_assert (! BIG_STRING_FULLSIZE_P (fullsize)); /* Just skip it if it isn't marked. */ if (! MARKED_RECORD_HEADER_P (&(string->lheader))) { from_pos += fullsize; continue; } /* If it won't fit in what's left of TO_SB, close TO_SB out and go on to the next string_chars_block. We know that TO_SB cannot advance past FROM_SB here since FROM_SB is large enough to currently contain this string. */ if ((to_pos + fullsize) > countof (to_sb->string_chars)) { to_sb->pos = to_pos; to_sb = to_sb->next; to_pos = 0; } /* Compute new address of this string and update TO_POS for the space being used. */ to_s_chars = (struct string_chars *) &(to_sb->string_chars[to_pos]); /* Copy the string_chars to the new place. */ if (from_s_chars != to_s_chars) memmove (to_s_chars, from_s_chars, fullsize); /* Relocate FROM_S_CHARS's reference */ set_string_data (string, &(to_s_chars->chars[0])); from_pos += fullsize; to_pos += fullsize; } } /* Set current to the last string chars block still used and free any that follow. */ { struct string_chars_block *victim; for (victim = to_sb->next; victim; ) { struct string_chars_block *next = victim->next; xfree (victim); victim = next; } current_string_chars_block = to_sb; current_string_chars_block->pos = to_pos; current_string_chars_block->next = 0; } } #if 1 /* Hack to debug missing purecopy's */ static int debug_string_purity; static void debug_string_purity_print (Lisp_String *p) { Charcount i; Charcount s = string_char_length (p); stderr_out ("\""); for (i = 0; i < s; i++) { Emchar ch = string_char (p, i); if (ch < 32 || ch >= 126) stderr_out ("\\%03o", ch); else if (ch == '\\' || ch == '\"') stderr_out ("\\%c", ch); else stderr_out ("%c", ch); } stderr_out ("\"\n"); } #endif /* 1 */ static void sweep_strings (void) { int num_small_used = 0, num_small_bytes = 0, num_bytes = 0; int debug = debug_string_purity; #define UNMARK_string(ptr) do { \ Lisp_String *p = (ptr); \ size_t size = string_length (p); \ UNMARK_RECORD_HEADER (&(p->lheader)); \ num_bytes += size; \ if (!BIG_STRING_SIZE_P (size)) \ { \ num_small_bytes += size; \ num_small_used++; \ } \ if (debug) \ debug_string_purity_print (p); \ } while (0) #define ADDITIONAL_FREE_string(ptr) do { \ size_t size = string_length (ptr); \ if (BIG_STRING_SIZE_P (size)) \ xfree (ptr->data); \ } while (0) SWEEP_FIXED_TYPE_BLOCK (string, Lisp_String); gc_count_num_short_string_in_use = num_small_used; gc_count_string_total_size = num_bytes; gc_count_short_string_total_size = num_small_bytes; } /* I hate duplicating all this crap! */ int marked_p (Lisp_Object obj) { /* Checks we used to perform. */ /* if (EQ (obj, Qnull_pointer)) return 1; */ /* if (!POINTER_TYPE_P (XGCTYPE (obj))) return 1; */ /* if (PURIFIED (XPNTR (obj))) return 1; */ if (XTYPE (obj) == Lisp_Type_Record) { struct lrecord_header *lheader = XRECORD_LHEADER (obj); GC_CHECK_LHEADER_INVARIANTS (lheader); return MARKED_RECORD_HEADER_P (lheader); } return 1; } static void gc_sweep (void) { /* Free all unmarked records. Do this at the very beginning, before anything else, so that the finalize methods can safely examine items in the objects. sweep_lcrecords_1() makes sure to call all the finalize methods *before* freeing anything, to complete the safety. */ { int ignored; sweep_lcrecords_1 (&all_lcrecords, &ignored); } compact_string_chars (); /* Finalize methods below (called through the ADDITIONAL_FREE_foo macros) must be *extremely* careful to make sure they're not referencing freed objects. The only two existing finalize methods (for strings and markers) pass muster -- the string finalizer doesn't look at anything but its own specially- created block, and the marker finalizer only looks at live buffers (which will never be freed) and at the markers before and after it in the chain (which, by induction, will never be freed because if so, they would have already removed themselves from the chain). */ /* Put all unmarked strings on free list, free'ing the string chars of large unmarked strings */ sweep_strings (); /* Put all unmarked conses on free list */ sweep_conses (); /* Free all unmarked bit vectors */ sweep_bit_vectors_1 (&all_bit_vectors, &gc_count_num_bit_vector_used, &gc_count_bit_vector_total_size, &gc_count_bit_vector_storage); /* Free all unmarked compiled-function objects */ sweep_compiled_functions (); #ifdef LISP_FLOAT_TYPE /* Put all unmarked floats on free list */ sweep_floats (); #endif /* Put all unmarked symbols on free list */ sweep_symbols (); /* Put all unmarked extents on free list */ sweep_extents (); /* Put all unmarked markers on free list. Dechain each one first from the buffer into which it points. */ sweep_markers (); sweep_events (); #ifdef PDUMP pdump_objects_unmark (); #endif } /* Clearing for disksave. */ void disksave_object_finalization (void) { /* It's important that certain information from the environment not get dumped with the executable (pathnames, environment variables, etc.). To make it easier to tell when this has happened with strings(1) we clear some known-to-be-garbage blocks of memory, so that leftover results of old evaluation don't look like potential problems. But first we set some notable variables to nil and do one more GC, to turn those strings into garbage. */ /* Yeah, this list is pretty ad-hoc... */ Vprocess_environment = Qnil; Vexec_directory = Qnil; Vdata_directory = Qnil; Vsite_directory = Qnil; Vdoc_directory = Qnil; Vconfigure_info_directory = Qnil; Vexec_path = Qnil; Vload_path = Qnil; /* Vdump_load_path = Qnil; */ /* Release hash tables for locate_file */ Flocate_file_clear_hashing (Qt); uncache_home_directory(); #if defined(LOADHIST) && !(defined(LOADHIST_DUMPED) || \ defined(LOADHIST_BUILTIN)) Vload_history = Qnil; #endif Vshell_file_name = Qnil; garbage_collect_1 (); /* Run the disksave finalization methods of all live objects. */ disksave_object_finalization_1 (); /* Zero out the uninitialized (really, unused) part of the containers for the live strings. */ { struct string_chars_block *scb; for (scb = first_string_chars_block; scb; scb = scb->next) { int count = sizeof (scb->string_chars) - scb->pos; assert (count >= 0 && count < STRING_CHARS_BLOCK_SIZE); if (count != 0) { /* from the block's fill ptr to the end */ memset ((scb->string_chars + scb->pos), 0, count); } } } /* There, that ought to be enough... */ } Lisp_Object restore_gc_inhibit (Lisp_Object val) { gc_currently_forbidden = XINT (val); return val; } /* Maybe we want to use this when doing a "panic" gc after memory_full()? */ static int gc_hooks_inhibited; struct post_gc_action { void (*fun) (void *); void *arg; }; typedef struct post_gc_action post_gc_action; typedef struct { Dynarr_declare (post_gc_action); } post_gc_action_dynarr; static post_gc_action_dynarr *post_gc_actions; /* Register an action to be called at the end of GC. gc_in_progress is 0 when this is called. This is used when it is discovered that an action needs to be taken, but it's during GC, so it's not safe. (e.g. in a finalize method.) As a general rule, do not use Lisp objects here. And NEVER signal an error. */ void register_post_gc_action (void (*fun) (void *), void *arg) { post_gc_action action; if (!post_gc_actions) post_gc_actions = Dynarr_new (post_gc_action); action.fun = fun; action.arg = arg; Dynarr_add (post_gc_actions, action); } static void run_post_gc_actions (void) { int i; if (post_gc_actions) { for (i = 0; i < Dynarr_length (post_gc_actions); i++) { post_gc_action action = Dynarr_at (post_gc_actions, i); (action.fun) (action.arg); } Dynarr_reset (post_gc_actions); } } void garbage_collect_1 (void) { #if MAX_SAVE_STACK > 0 char stack_top_variable; extern char *stack_bottom; #endif struct frame *f; int speccount; int cursor_changed; Lisp_Object pre_gc_cursor; struct gcpro gcpro1; if (gc_in_progress || gc_currently_forbidden || in_display || preparing_for_armageddon) return; /* We used to call selected_frame() here. The following functions cannot be called inside GC so we move to after the above tests. */ { Lisp_Object frame; Lisp_Object device = Fselected_device (Qnil); if (NILP (device)) /* Could happen during startup, eg. if always_gc */ return; frame = DEVICE_SELECTED_FRAME (XDEVICE (device)); if (NILP (frame)) signal_simple_error ("No frames exist on device", device); f = XFRAME (frame); } pre_gc_cursor = Qnil; cursor_changed = 0; GCPRO1 (pre_gc_cursor); /* Very important to prevent GC during any of the following stuff that might run Lisp code; otherwise, we'll likely have infinite GC recursion. */ speccount = specpdl_depth (); record_unwind_protect (restore_gc_inhibit, make_int (gc_currently_forbidden)); gc_currently_forbidden = 1; if (!gc_hooks_inhibited) run_hook_trapping_errors ("Error in pre-gc-hook", Qpre_gc_hook); /* Now show the GC cursor/message. */ if (!noninteractive) { if (FRAME_WIN_P (f)) { Lisp_Object frame = make_frame (f); Lisp_Object cursor = glyph_image_instance (Vgc_pointer_glyph, FRAME_SELECTED_WINDOW (f), ERROR_ME_NOT, 1); pre_gc_cursor = f->pointer; if (POINTER_IMAGE_INSTANCEP (cursor) /* don't change if we don't know how to change back. */ && POINTER_IMAGE_INSTANCEP (pre_gc_cursor)) { cursor_changed = 1; Fset_frame_pointer (frame, cursor); } } /* Don't print messages to the stream device. */ if (!cursor_changed && !FRAME_STREAM_P (f)) { char *msg = (STRINGP (Vgc_message) ? GETTEXT ((char *) XSTRING_DATA (Vgc_message)) : 0); Lisp_Object args[2], whole_msg; args[0] = build_string (msg ? msg : GETTEXT ((const char *) gc_default_message)); args[1] = build_string ("..."); whole_msg = Fconcat (2, args); echo_area_message (f, (Bufbyte *) 0, whole_msg, 0, -1, Qgarbage_collecting); } } /***** Now we actually start the garbage collection. */ gc_in_progress = 1; inhibit_non_essential_printing_operations = 1; gc_generation_number[0]++; #if MAX_SAVE_STACK > 0 /* Save a copy of the contents of the stack, for debugging. */ if (!purify_flag) { /* Static buffer in which we save a copy of the C stack at each GC. */ static char *stack_copy; static size_t stack_copy_size; ptrdiff_t stack_diff = &stack_top_variable - stack_bottom; size_t stack_size = (stack_diff > 0 ? stack_diff : -stack_diff); if (stack_size < MAX_SAVE_STACK) { if (stack_copy_size < stack_size) { stack_copy = (char *) xrealloc (stack_copy, stack_size); stack_copy_size = stack_size; } memcpy (stack_copy, stack_diff > 0 ? stack_bottom : &stack_top_variable, stack_size); } } #endif /* MAX_SAVE_STACK > 0 */ /* Do some totally ad-hoc resource clearing. */ /* #### generalize this? */ clear_event_resource (); cleanup_specifiers (); /* Mark all the special slots that serve as the roots of accessibility. */ { /* staticpro() */ Lisp_Object **p = Dynarr_begin (staticpros); size_t count; for (count = Dynarr_length (staticpros); count; count--) mark_object (**p++); } { /* staticpro_nodump() */ Lisp_Object **p = Dynarr_begin (staticpros_nodump); size_t count; for (count = Dynarr_length (staticpros_nodump); count; count--) mark_object (**p++); } { /* GCPRO() */ struct gcpro *tail; int i; for (tail = gcprolist; tail; tail = tail->next) for (i = 0; i < tail->nvars; i++) mark_object (tail->var[i]); } { /* specbind() */ struct specbinding *bind; for (bind = specpdl; bind != specpdl_ptr; bind++) { mark_object (bind->symbol); mark_object (bind->old_value); } } { struct catchtag *catch; for (catch = catchlist; catch; catch = catch->next) { mark_object (catch->tag); mark_object (catch->val); } } { struct backtrace *backlist; for (backlist = backtrace_list; backlist; backlist = backlist->next) { int nargs = backlist->nargs; int i; mark_object (*backlist->function); if (nargs < 0 /* nargs == UNEVALLED || nargs == MANY */) mark_object (backlist->args[0]); else for (i = 0; i < nargs; i++) mark_object (backlist->args[i]); } } mark_redisplay (); mark_profiling_info (); /* OK, now do the after-mark stuff. This is for things that are only marked when something else is marked (e.g. weak hash tables). There may be complex dependencies between such objects -- e.g. a weak hash table might be unmarked, but after processing a later weak hash table, the former one might get marked. So we have to iterate until nothing more gets marked. */ while (finish_marking_weak_hash_tables () > 0 || finish_marking_weak_lists () > 0) ; /* And prune (this needs to be called after everything else has been marked and before we do any sweeping). */ /* #### this is somewhat ad-hoc and should probably be an object method */ prune_weak_hash_tables (); prune_weak_lists (); prune_specifiers (); prune_syntax_tables (); gc_sweep (); consing_since_gc = 0; #ifndef DEBUG_XEMACS /* Allow you to set it really fucking low if you really want ... */ if (gc_cons_threshold < 10000) gc_cons_threshold = 10000; #endif inhibit_non_essential_printing_operations = 0; gc_in_progress = 0; run_post_gc_actions (); /******* End of garbage collection ********/ run_hook_trapping_errors ("Error in post-gc-hook", Qpost_gc_hook); /* Now remove the GC cursor/message */ if (!noninteractive) { if (cursor_changed) Fset_frame_pointer (make_frame (f), pre_gc_cursor); else if (!FRAME_STREAM_P (f)) { char *msg = (STRINGP (Vgc_message) ? GETTEXT ((char *) XSTRING_DATA (Vgc_message)) : 0); /* Show "...done" only if the echo area would otherwise be empty. */ if (NILP (clear_echo_area (selected_frame (), Qgarbage_collecting, 0))) { Lisp_Object args[2], whole_msg; args[0] = build_string (msg ? msg : GETTEXT ((const char *) gc_default_message)); args[1] = build_string ("... done"); whole_msg = Fconcat (2, args); echo_area_message (selected_frame (), (Bufbyte *) 0, whole_msg, 0, -1, Qgarbage_collecting); } } } /* now stop inhibiting GC */ unbind_to (speccount, Qnil); if (!breathing_space) { breathing_space = malloc (4096 - MALLOC_OVERHEAD); } UNGCPRO; return; } /* Debugging aids. */ static Lisp_Object gc_plist_hack (const char *name, int value, Lisp_Object tail) { /* C doesn't have local functions (or closures, or GC, or readable syntax, or portable numeric datatypes, or bit-vectors, or characters, or arrays, or exceptions, or ...) */ return cons3 (intern (name), make_int (value), tail); } #define HACK_O_MATIC(type, name, pl) do { \ int s = 0; \ struct type##_block *x = current_##type##_block; \ while (x) { s += sizeof (*x) + MALLOC_OVERHEAD; x = x->prev; } \ (pl) = gc_plist_hack ((name), s, (pl)); \ } while (0) DEFUN ("garbage-collect", Fgarbage_collect, 0, 0, "", /* Reclaim storage for Lisp objects no longer needed. Return info on amount of space in use: ((USED-CONSES . FREE-CONSES) (USED-SYMS . FREE-SYMS) (USED-MARKERS . FREE-MARKERS) USED-STRING-CHARS USED-VECTOR-SLOTS PLIST) where `PLIST' is a list of alternating keyword/value pairs providing more detailed information. Garbage collection happens automatically if you cons more than `gc-cons-threshold' bytes of Lisp data since previous garbage collection. */ ()) { Lisp_Object pl = Qnil; unsigned int i; int gc_count_vector_total_size = 0; garbage_collect_1 (); for (i = 0; i < lrecord_type_count; i++) { if (lcrecord_stats[i].bytes_in_use != 0 || lcrecord_stats[i].bytes_freed != 0 || lcrecord_stats[i].instances_on_free_list != 0) { char buf [255]; const char *name = lrecord_implementations_table[i]->name; int len = strlen (name); /* save this for the FSFmacs-compatible part of the summary */ if (i == lrecord_type_vector) gc_count_vector_total_size = lcrecord_stats[i].bytes_in_use + lcrecord_stats[i].bytes_freed; sprintf (buf, "%s-storage", name); pl = gc_plist_hack (buf, lcrecord_stats[i].bytes_in_use, pl); /* Okay, simple pluralization check for `symbol-value-varalias' */ if (name[len-1] == 's') sprintf (buf, "%ses-freed", name); else sprintf (buf, "%ss-freed", name); if (lcrecord_stats[i].instances_freed != 0) pl = gc_plist_hack (buf, lcrecord_stats[i].instances_freed, pl); if (name[len-1] == 's') sprintf (buf, "%ses-on-free-list", name); else sprintf (buf, "%ss-on-free-list", name); if (lcrecord_stats[i].instances_on_free_list != 0) pl = gc_plist_hack (buf, lcrecord_stats[i].instances_on_free_list, pl); if (name[len-1] == 's') sprintf (buf, "%ses-used", name); else sprintf (buf, "%ss-used", name); pl = gc_plist_hack (buf, lcrecord_stats[i].instances_in_use, pl); } } HACK_O_MATIC (extent, "extent-storage", pl); pl = gc_plist_hack ("extents-free", gc_count_num_extent_freelist, pl); pl = gc_plist_hack ("extents-used", gc_count_num_extent_in_use, pl); HACK_O_MATIC (event, "event-storage", pl); pl = gc_plist_hack ("events-free", gc_count_num_event_freelist, pl); pl = gc_plist_hack ("events-used", gc_count_num_event_in_use, pl); HACK_O_MATIC (marker, "marker-storage", pl); pl = gc_plist_hack ("markers-free", gc_count_num_marker_freelist, pl); pl = gc_plist_hack ("markers-used", gc_count_num_marker_in_use, pl); #ifdef LISP_FLOAT_TYPE HACK_O_MATIC (float, "float-storage", pl); pl = gc_plist_hack ("floats-free", gc_count_num_float_freelist, pl); pl = gc_plist_hack ("floats-used", gc_count_num_float_in_use, pl); #endif /* LISP_FLOAT_TYPE */ HACK_O_MATIC (string, "string-header-storage", pl); pl = gc_plist_hack ("long-strings-total-length", gc_count_string_total_size - gc_count_short_string_total_size, pl); HACK_O_MATIC (string_chars, "short-string-storage", pl); pl = gc_plist_hack ("short-strings-total-length", gc_count_short_string_total_size, pl); pl = gc_plist_hack ("strings-free", gc_count_num_string_freelist, pl); pl = gc_plist_hack ("long-strings-used", gc_count_num_string_in_use - gc_count_num_short_string_in_use, pl); pl = gc_plist_hack ("short-strings-used", gc_count_num_short_string_in_use, pl); HACK_O_MATIC (compiled_function, "compiled-function-storage", pl); pl = gc_plist_hack ("compiled-functions-free", gc_count_num_compiled_function_freelist, pl); pl = gc_plist_hack ("compiled-functions-used", gc_count_num_compiled_function_in_use, pl); pl = gc_plist_hack ("bit-vector-storage", gc_count_bit_vector_storage, pl); pl = gc_plist_hack ("bit-vectors-total-length", gc_count_bit_vector_total_size, pl); pl = gc_plist_hack ("bit-vectors-used", gc_count_num_bit_vector_used, pl); HACK_O_MATIC (symbol, "symbol-storage", pl); pl = gc_plist_hack ("symbols-free", gc_count_num_symbol_freelist, pl); pl = gc_plist_hack ("symbols-used", gc_count_num_symbol_in_use, pl); HACK_O_MATIC (cons, "cons-storage", pl); pl = gc_plist_hack ("conses-free", gc_count_num_cons_freelist, pl); pl = gc_plist_hack ("conses-used", gc_count_num_cons_in_use, pl); /* The things we do for backwards-compatibility */ return list6 (Fcons (make_int (gc_count_num_cons_in_use), make_int (gc_count_num_cons_freelist)), Fcons (make_int (gc_count_num_symbol_in_use), make_int (gc_count_num_symbol_freelist)), Fcons (make_int (gc_count_num_marker_in_use), make_int (gc_count_num_marker_freelist)), make_int (gc_count_string_total_size), make_int (gc_count_vector_total_size), pl); } #undef HACK_O_MATIC DEFUN ("consing-since-gc", Fconsing_since_gc, 0, 0, "", /* Return the number of bytes consed since the last garbage collection. \"Consed\" is a misnomer in that this actually counts allocation of all different kinds of objects, not just conses. If this value exceeds `gc-cons-threshold', a garbage collection happens. */ ()) { return make_int (consing_since_gc); } #if 0 DEFUN ("memory-limit", Fmemory_limit, 0, 0, 0, /* Return the address of the last byte Emacs has allocated, divided by 1024. This may be helpful in debugging Emacs's memory usage. The value is divided by 1024 to make sure it will fit in a lisp integer. */ ()) { return make_int ((EMACS_INT) sbrk (0) / 1024); } #endif int object_dead_p (Lisp_Object obj) { return ((BUFFERP (obj) && !BUFFER_LIVE_P (XBUFFER (obj))) || (FRAMEP (obj) && !FRAME_LIVE_P (XFRAME (obj))) || (WINDOWP (obj) && !WINDOW_LIVE_P (XWINDOW (obj))) || (DEVICEP (obj) && !DEVICE_LIVE_P (XDEVICE (obj))) || (CONSOLEP (obj) && !CONSOLE_LIVE_P (XCONSOLE (obj))) || (EVENTP (obj) && !EVENT_LIVE_P (XEVENT (obj))) || (EXTENTP (obj) && !EXTENT_LIVE_P (XEXTENT (obj)))); } #ifdef MEMORY_USAGE_STATS /* Attempt to determine the actual amount of space that is used for the block allocated starting at PTR, supposedly of size "CLAIMED_SIZE". It seems that the following holds: 1. When using the old allocator (malloc.c): -- blocks are always allocated in chunks of powers of two. For each block, there is an overhead of 8 bytes if rcheck is not defined, 20 bytes if it is defined. In other words, a one-byte allocation needs 8 bytes of overhead for a total of 9 bytes, and needs to have 16 bytes of memory chunked out for it. 2. When using the new allocator (gmalloc.c): -- blocks are always allocated in chunks of powers of two up to 4096 bytes. Larger blocks are allocated in chunks of an integral multiple of 4096 bytes. The minimum block size is 2*sizeof (void *), or 16 bytes if SUNOS_LOCALTIME_BUG is defined. There is no per-block overhead, but there is an overhead of 3*sizeof (size_t) for each 4096 bytes allocated. 3. When using the system malloc, anything goes, but they are generally slower and more space-efficient than the GNU allocators. One possibly reasonable assumption to make for want of better data is that sizeof (void *), or maybe 2 * sizeof (void *), is required as overhead and that blocks are allocated in the minimum required size except that some minimum block size is imposed (e.g. 16 bytes). */ size_t malloced_storage_size (void *ptr, size_t claimed_size, struct overhead_stats *stats) { size_t orig_claimed_size = claimed_size; #ifdef GNU_MALLOC if (claimed_size < 2 * sizeof (void *)) claimed_size = 2 * sizeof (void *); # ifdef SUNOS_LOCALTIME_BUG if (claimed_size < 16) claimed_size = 16; # endif if (claimed_size < 4096) { int log = 1; /* compute the log base two, more or less, then use it to compute the block size needed. */ claimed_size--; /* It's big, it's heavy, it's wood! */ while ((claimed_size /= 2) != 0) ++log; claimed_size = 1; /* It's better than bad, it's good! */ while (log > 0) { claimed_size *= 2; log--; } /* We have to come up with some average about the amount of blocks used. */ if ((size_t) (rand () & 4095) < claimed_size) claimed_size += 3 * sizeof (void *); } else { claimed_size += 4095; claimed_size &= ~4095; claimed_size += (claimed_size / 4096) * 3 * sizeof (size_t); } #elif defined (SYSTEM_MALLOC) if (claimed_size < 16) claimed_size = 16; claimed_size += 2 * sizeof (void *); #else /* old GNU allocator */ # ifdef rcheck /* #### may not be defined here */ claimed_size += 20; # else claimed_size += 8; # endif { int log = 1; /* compute the log base two, more or less, then use it to compute the block size needed. */ claimed_size--; /* It's big, it's heavy, it's wood! */ while ((claimed_size /= 2) != 0) ++log; claimed_size = 1; /* It's better than bad, it's good! */ while (log > 0) { claimed_size *= 2; log--; } } #endif /* old GNU allocator */ if (stats) { stats->was_requested += orig_claimed_size; stats->malloc_overhead += claimed_size - orig_claimed_size; } return claimed_size; } size_t fixed_type_block_overhead (size_t size) { size_t per_block = TYPE_ALLOC_SIZE (cons, unsigned char); size_t overhead = 0; size_t storage_size = malloced_storage_size (0, per_block, 0); while (size >= per_block) { size -= per_block; overhead += sizeof (void *) + per_block - storage_size; } if (rand () % per_block < size) overhead += sizeof (void *) + per_block - storage_size; return overhead; } #endif /* MEMORY_USAGE_STATS */ /* Initialization */ void reinit_alloc_once_early (void) { gc_generation_number[0] = 0; breathing_space = 0; XSETINT (all_bit_vectors, 0); /* Qzero may not be set yet. */ XSETINT (Vgc_message, 0); all_lcrecords = 0; ignore_malloc_warnings = 1; #ifdef DOUG_LEA_MALLOC mallopt (M_TRIM_THRESHOLD, 128*1024); /* trim threshold */ mallopt (M_MMAP_THRESHOLD, 64*1024); /* mmap threshold */ #if 1 /* Moved to emacs.c */ mallopt (M_MMAP_MAX, 0); /* max. number of mmap'ed areas */ #endif #endif init_string_alloc (); init_string_chars_alloc (); init_cons_alloc (); init_symbol_alloc (); init_compiled_function_alloc (); #ifdef LISP_FLOAT_TYPE init_float_alloc (); #endif /* LISP_FLOAT_TYPE */ init_marker_alloc (); init_extent_alloc (); init_event_alloc (); ignore_malloc_warnings = 0; if (staticpros_nodump) Dynarr_free (staticpros_nodump); staticpros_nodump = Dynarr_new2 (Lisp_Object_ptr_dynarr, Lisp_Object *); Dynarr_resize (staticpros_nodump, 100); /* merely a small optimization */ consing_since_gc = 0; #if 1 gc_cons_threshold = 500000; /* XEmacs change */ #else gc_cons_threshold = 15000; /* debugging */ #endif lrecord_uid_counter = 259; debug_string_purity = 0; gcprolist = 0; gc_currently_forbidden = 0; gc_hooks_inhibited = 0; #ifdef ERROR_CHECK_TYPECHECK ERROR_ME.really_unlikely_name_to_have_accidentally_in_a_non_errb_structure = 666; ERROR_ME_NOT. really_unlikely_name_to_have_accidentally_in_a_non_errb_structure = 42; ERROR_ME_WARN. really_unlikely_name_to_have_accidentally_in_a_non_errb_structure = 3333632; #endif /* ERROR_CHECK_TYPECHECK */ } void init_alloc_once_early (void) { reinit_alloc_once_early (); { int i; for (i = 0; i < countof (lrecord_implementations_table); i++) lrecord_implementations_table[i] = 0; } INIT_LRECORD_IMPLEMENTATION (cons); INIT_LRECORD_IMPLEMENTATION (vector); INIT_LRECORD_IMPLEMENTATION (string); INIT_LRECORD_IMPLEMENTATION (lcrecord_list); staticpros = Dynarr_new2 (Lisp_Object_ptr_dynarr, Lisp_Object *); Dynarr_resize (staticpros, 1410); /* merely a small optimization */ dump_add_root_struct_ptr (&staticpros, &staticpros_description); } void reinit_alloc (void) { gcprolist = 0; } void syms_of_alloc (void) { DEFSYMBOL (Qpre_gc_hook); DEFSYMBOL (Qpost_gc_hook); DEFSYMBOL (Qgarbage_collecting); DEFSUBR (Fcons); DEFSUBR (Flist); DEFSUBR (Fvector); DEFSUBR (Fbit_vector); DEFSUBR (Fmake_byte_code); DEFSUBR (Fmake_list); DEFSUBR (Fmake_vector); DEFSUBR (Fmake_bit_vector); DEFSUBR (Fmake_string); DEFSUBR (Fstring); DEFSUBR (Fmake_symbol); DEFSUBR (Fmake_marker); DEFSUBR (Fpurecopy); DEFSUBR (Fgarbage_collect); #if 0 DEFSUBR (Fmemory_limit); #endif DEFSUBR (Fconsing_since_gc); } void vars_of_alloc (void) { DEFVAR_INT ("gc-cons-threshold", &gc_cons_threshold /* *Number of bytes of consing between garbage collections. \"Consing\" is a misnomer in that this actually counts allocation of all different kinds of objects, not just conses. Garbage collection can happen automatically once this many bytes have been allocated since the last garbage collection. All data types count. Garbage collection happens automatically when `eval' or `funcall' are called. (Note that `funcall' is called implicitly as part of evaluation.) By binding this temporarily to a large number, you can effectively prevent garbage collection during a part of the program. See also `consing-since-gc'. */ ); #ifdef DEBUG_XEMACS DEFVAR_INT ("debug-allocation", &debug_allocation /* If non-zero, print out information to stderr about all objects allocated. See also `debug-allocation-backtrace-length'. */ ); debug_allocation = 0; DEFVAR_INT ("debug-allocation-backtrace-length", &debug_allocation_backtrace_length /* Length (in stack frames) of short backtrace printed out by `debug-allocation'. */ ); debug_allocation_backtrace_length = 2; #endif DEFVAR_BOOL ("purify-flag", &purify_flag /* Non-nil means loading Lisp code in order to dump an executable. This means that certain objects should be allocated in readonly space. */ ); DEFVAR_LISP ("pre-gc-hook", &Vpre_gc_hook /* Function or functions to be run just before each garbage collection. Interrupts, garbage collection, and errors are inhibited while this hook runs, so be extremely careful in what you add here. In particular, avoid consing, and do not interact with the user. */ ); Vpre_gc_hook = Qnil; DEFVAR_LISP ("post-gc-hook", &Vpost_gc_hook /* Function or functions to be run just after each garbage collection. Interrupts, garbage collection, and errors are inhibited while this hook runs, so be extremely careful in what you add here. In particular, avoid consing, and do not interact with the user. */ ); Vpost_gc_hook = Qnil; DEFVAR_LISP ("gc-message", &Vgc_message /* String to print to indicate that a garbage collection is in progress. This is printed in the echo area. If the selected frame is on a window system and `gc-pointer-glyph' specifies a value (i.e. a pointer image instance) in the domain of the selected frame, the mouse pointer will change instead of this message being printed. */ ); Vgc_message = build_string (gc_default_message); DEFVAR_LISP ("gc-pointer-glyph", &Vgc_pointer_glyph /* Pointer glyph used to indicate that a garbage collection is in progress. If the selected window is on a window system and this glyph specifies a value (i.e. a pointer image instance) in the domain of the selected window, the pointer will be changed as specified during garbage collection. Otherwise, a message will be printed in the echo area, as controlled by `gc-message'. */ ); } void complex_vars_of_alloc (void) { Vgc_pointer_glyph = Fmake_glyph_internal (Qpointer); }