gcmodule.c 48.1 KB
Newer Older
1
/*
2

3 4 5
  Reference Cycle Garbage Collection
  ==================================

6
  Neil Schemenauer <nas@arctrix.com>
7 8 9 10

  Based on a post on the python-dev list.  Ideas from Guido van Rossum,
  Eric Tiedemann, and various others.

11
  http://www.arctrix.com/nas/python/gc/
12 13 14 15 16 17 18 19

  The following mailing list threads provide a historical perspective on
  the design of this module.  Note that a fair amount of refinement has
  occurred since those discussions.

  http://mail.python.org/pipermail/python-dev/2000-March/002385.html
  http://mail.python.org/pipermail/python-dev/2000-March/002434.html
  http://mail.python.org/pipermail/python-dev/2000-March/002497.html
20 21 22 23 24 25 26

  For a highlevel view of the collection process, read the collect
  function.

*/

#include "Python.h"
27
#include "frameobject.h"        /* for PyFrame_ClearFreeList */
28

29 30 31 32 33 34
/* Get an object's GC head */
#define AS_GC(o) ((PyGC_Head *)(o)-1)

/* Get the object given the GC head */
#define FROM_GC(g) ((PyObject *)(((PyGC_Head *)g)+1))

35 36
/*** Global GC state ***/

37
struct gc_generation {
38 39 40 41
    PyGC_Head head;
    int threshold; /* collection threshold */
    int count; /* count of allocations or collections of younger
                  generations */
42 43 44 45 46
};

#define NUM_GENERATIONS 3
#define GEN_HEAD(n) (&generations[n].head)

47
/* linked lists of container objects */
48
static struct gc_generation generations[NUM_GENERATIONS] = {
49 50 51 52
    /* PyGC_Head,                               threshold,      count */
    {{{GEN_HEAD(0), GEN_HEAD(0), 0}},           700,            0},
    {{{GEN_HEAD(1), GEN_HEAD(1), 0}},           10,             0},
    {{{GEN_HEAD(2), GEN_HEAD(2), 0}},           10,             0},
53
};
54

55
PyGC_Head *_PyGC_generation0 = GEN_HEAD(0);
56

57
static int enabled = 1; /* automatic collection enabled? */
58

59
/* true if we are currently running the collector */
60
static int collecting = 0;
61

62
/* list of uncollectable objects */
63
static PyObject *garbage = NULL;
64 65

/* Python string to use if unhandled exception occurs */
66
static PyObject *gc_str = NULL;
67

68 69
/* Python string used to look for __del__ attribute. */
static PyObject *delstr = NULL;
Jeremy Hylton's avatar
Jeremy Hylton committed
70

71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93
/* This is the number of objects who survived the last full collection. It
   approximates the number of long lived objects tracked by the GC.

   (by "full collection", we mean a collection of the oldest generation).
*/
static Py_ssize_t long_lived_total = 0;

/* This is the number of objects who survived all "non-full" collections,
   and are awaiting to undergo a full collection for the first time.

*/
static Py_ssize_t long_lived_pending = 0;

/*
   NOTE: about the counting of long-lived objects.

   To limit the cost of garbage collection, there are two strategies;
     - make each collection faster, e.g. by scanning fewer objects
     - do less collections
   This heuristic is about the latter strategy.

   In addition to the various configurable thresholds, we only trigger a
   full collection if the ratio
94
    long_lived_pending / long_lived_total
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
   is above a given value (hardwired to 25%).

   The reason is that, while "non-full" collections (i.e., collections of
   the young and middle generations) will always examine roughly the same
   number of objects -- determined by the aforementioned thresholds --,
   the cost of a full collection is proportional to the total number of
   long-lived objects, which is virtually unbounded.

   Indeed, it has been remarked that doing a full collection every
   <constant number> of object creations entails a dramatic performance
   degradation in workloads which consist in creating and storing lots of
   long-lived objects (e.g. building a large list of GC-tracked objects would
   show quadratic performance, instead of linear as expected: see issue #4074).

   Using the above ratio, instead, yields amortized linear performance in
   the total number of objects (the effect of which can be summarized
   thusly: "each full garbage collection is more and more costly as the
   number of objects grows, but we do fewer and fewer of them").

   This heuristic was suggested by Martin von Löwis on python-dev in
   June 2008. His original analysis and proposal can be found at:
116
    http://mail.python.org/pipermail/python-dev/2008-June/080579.html
117 118 119
*/


120
/* set for debugging information */
121 122 123 124 125 126 127
#define DEBUG_STATS             (1<<0) /* print collection statistics */
#define DEBUG_COLLECTABLE       (1<<1) /* print collectable objects */
#define DEBUG_UNCOLLECTABLE     (1<<2) /* print uncollectable objects */
#define DEBUG_SAVEALL           (1<<5) /* save all garbage in gc.garbage */
#define DEBUG_LEAK              DEBUG_COLLECTABLE | \
                DEBUG_UNCOLLECTABLE | \
                DEBUG_SAVEALL
128
static int debug;
129
static PyObject *tmod = NULL;
130

131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
/*--------------------------------------------------------------------------
gc_refs values.

Between collections, every gc'ed object has one of two gc_refs values:

GC_UNTRACKED
    The initial state; objects returned by PyObject_GC_Malloc are in this
    state.  The object doesn't live in any generation list, and its
    tp_traverse slot must not be called.

GC_REACHABLE
    The object lives in some generation list, and its tp_traverse is safe to
    call.  An object transitions to GC_REACHABLE when PyObject_GC_Track
    is called.

During a collection, gc_refs can temporarily take on other states:

>= 0
    At the start of a collection, update_refs() copies the true refcount
    to gc_refs, for each object in the generation being collected.
    subtract_refs() then adjusts gc_refs so that it equals the number of
    times an object is referenced directly from outside the generation
    being collected.
Martin v. Löwis's avatar
Martin v. Löwis committed
154
    gc_refs remains >= 0 throughout these steps.
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169

GC_TENTATIVELY_UNREACHABLE
    move_unreachable() then moves objects not reachable (whether directly or
    indirectly) from outside the generation into an "unreachable" set.
    Objects that are found to be reachable have gc_refs set to GC_REACHABLE
    again.  Objects that are found to be unreachable have gc_refs set to
    GC_TENTATIVELY_UNREACHABLE.  It's "tentatively" because the pass doing
    this can't be sure until it ends, and GC_TENTATIVELY_UNREACHABLE may
    transition back to GC_REACHABLE.

    Only objects with GC_TENTATIVELY_UNREACHABLE still set are candidates
    for collection.  If it's decided not to collect such an object (e.g.,
    it has a __del__ method), its gc_refs is restored to GC_REACHABLE again.
----------------------------------------------------------------------------
*/
170 171 172
#define GC_UNTRACKED                    _PyGC_REFS_UNTRACKED
#define GC_REACHABLE                    _PyGC_REFS_REACHABLE
#define GC_TENTATIVELY_UNREACHABLE      _PyGC_REFS_TENTATIVELY_UNREACHABLE
173

174
#define IS_TRACKED(o) ((AS_GC(o))->gc.gc_refs != GC_UNTRACKED)
175 176
#define IS_REACHABLE(o) ((AS_GC(o))->gc.gc_refs == GC_REACHABLE)
#define IS_TENTATIVELY_UNREACHABLE(o) ( \
177
    (AS_GC(o))->gc.gc_refs == GC_TENTATIVELY_UNREACHABLE)
178

179 180 181 182 183
/*** list functions ***/

static void
gc_list_init(PyGC_Head *list)
{
184 185
    list->gc.gc_prev = list;
    list->gc.gc_next = list;
186 187
}

188 189 190
static int
gc_list_is_empty(PyGC_Head *list)
{
191
    return (list->gc.gc_next == list);
192 193
}

Tim Peters's avatar
Tim Peters committed
194 195 196
#if 0
/* This became unused after gc_list_move() was introduced. */
/* Append `node` to `list`. */
197 198 199
static void
gc_list_append(PyGC_Head *node, PyGC_Head *list)
{
200 201 202 203
    node->gc.gc_next = list;
    node->gc.gc_prev = list->gc.gc_prev;
    node->gc.gc_prev->gc.gc_next = node;
    list->gc.gc_prev = node;
204
}
Tim Peters's avatar
Tim Peters committed
205
#endif
206

Tim Peters's avatar
Tim Peters committed
207
/* Remove `node` from the gc list it's currently in. */
208 209 210
static void
gc_list_remove(PyGC_Head *node)
{
211 212 213
    node->gc.gc_prev->gc.gc_next = node->gc.gc_next;
    node->gc.gc_next->gc.gc_prev = node->gc.gc_prev;
    node->gc.gc_next = NULL; /* object is not currently tracked */
214 215
}

Tim Peters's avatar
Tim Peters committed
216 217 218 219 220 221 222
/* Move `node` from the gc list it's currently in (which is not explicitly
 * named here) to the end of `list`.  This is semantically the same as
 * gc_list_remove(node) followed by gc_list_append(node, list).
 */
static void
gc_list_move(PyGC_Head *node, PyGC_Head *list)
{
223 224 225 226 227 228 229 230 231 232
    PyGC_Head *new_prev;
    PyGC_Head *current_prev = node->gc.gc_prev;
    PyGC_Head *current_next = node->gc.gc_next;
    /* Unlink from current list. */
    current_prev->gc.gc_next = current_next;
    current_next->gc.gc_prev = current_prev;
    /* Relink at end of new list. */
    new_prev = node->gc.gc_prev = list->gc.gc_prev;
    new_prev->gc.gc_next = list->gc.gc_prev = node;
    node->gc.gc_next = list;
Tim Peters's avatar
Tim Peters committed
233 234 235
}

/* append list `from` onto list `to`; `from` becomes an empty list */
236 237 238
static void
gc_list_merge(PyGC_Head *from, PyGC_Head *to)
{
239 240 241 242 243 244 245 246 247 248
    PyGC_Head *tail;
    assert(from != to);
    if (!gc_list_is_empty(from)) {
        tail = to->gc.gc_prev;
        tail->gc.gc_next = from->gc.gc_next;
        tail->gc.gc_next->gc.gc_prev = tail;
        to->gc.gc_prev = from->gc.gc_prev;
        to->gc.gc_prev->gc.gc_next = to;
    }
    gc_list_init(from);
249 250
}

251
static Py_ssize_t
252 253
gc_list_size(PyGC_Head *list)
{
254 255 256 257 258 259
    PyGC_Head *gc;
    Py_ssize_t n = 0;
    for (gc = list->gc.gc_next; gc != list; gc = gc->gc.gc_next) {
        n++;
    }
    return n;
260 261
}

262 263 264 265 266 267
/* Append objects in a GC list to a Python list.
 * Return 0 if all OK, < 0 if error (out of memory for list).
 */
static int
append_objects(PyObject *py_list, PyGC_Head *gc_list)
{
268 269 270 271 272 273 274 275 276 277
    PyGC_Head *gc;
    for (gc = gc_list->gc.gc_next; gc != gc_list; gc = gc->gc.gc_next) {
        PyObject *op = FROM_GC(gc);
        if (op != py_list) {
            if (PyList_Append(py_list, op)) {
                return -1; /* exception */
            }
        }
    }
    return 0;
278 279
}

280 281 282
/*** end of list stuff ***/


283 284 285
/* Set all gc_refs = ob_refcnt.  After this, gc_refs is > 0 for all objects
 * in containers, and is GC_REACHABLE for all tracked gc objects not in
 * containers.
286
 */
287 288 289
static void
update_refs(PyGC_Head *containers)
{
290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313
    PyGC_Head *gc = containers->gc.gc_next;
    for (; gc != containers; gc = gc->gc.gc_next) {
        assert(gc->gc.gc_refs == GC_REACHABLE);
        gc->gc.gc_refs = Py_REFCNT(FROM_GC(gc));
        /* Python's cyclic gc should never see an incoming refcount
         * of 0:  if something decref'ed to 0, it should have been
         * deallocated immediately at that time.
         * Possible cause (if the assert triggers):  a tp_dealloc
         * routine left a gc-aware object tracked during its teardown
         * phase, and did something-- or allowed something to happen --
         * that called back into Python.  gc can trigger then, and may
         * see the still-tracked dying object.  Before this assert
         * was added, such mistakes went on to allow gc to try to
         * delete the object again.  In a debug build, that caused
         * a mysterious segfault, when _Py_ForgetReference tried
         * to remove the object from the doubly-linked list of all
         * objects a second time.  In a release build, an actual
         * double deallocation occurred, which leads to corruption
         * of the allocator's internal bookkeeping pointers.  That's
         * so serious that maybe this should be a release-build
         * check instead of an assert?
         */
        assert(gc->gc.gc_refs != 0);
    }
314 315
}

316
/* A traversal callback for subtract_refs. */
317 318 319
static int
visit_decref(PyObject *op, void *data)
{
320 321 322 323 324 325 326 327 328 329 330 331
    assert(op != NULL);
    if (PyObject_IS_GC(op)) {
        PyGC_Head *gc = AS_GC(op);
        /* We're only interested in gc_refs for objects in the
         * generation being collected, which can be recognized
         * because only they have positive gc_refs.
         */
        assert(gc->gc.gc_refs != 0); /* else refcount was too small */
        if (gc->gc.gc_refs > 0)
            gc->gc.gc_refs--;
    }
    return 0;
332 333
}

334 335 336 337 338
/* Subtract internal references from gc_refs.  After this, gc_refs is >= 0
 * for all objects in containers, and is GC_REACHABLE for all tracked gc
 * objects not in containers.  The ones with gc_refs > 0 are directly
 * reachable from outside containers, and so can't be collected.
 */
339 340 341
static void
subtract_refs(PyGC_Head *containers)
{
342 343 344 345 346 347 348 349
    traverseproc traverse;
    PyGC_Head *gc = containers->gc.gc_next;
    for (; gc != containers; gc=gc->gc.gc_next) {
        traverse = Py_TYPE(FROM_GC(gc))->tp_traverse;
        (void) traverse(FROM_GC(gc),
                       (visitproc)visit_decref,
                       NULL);
    }
350 351
}

352
/* A traversal callback for move_unreachable. */
353
static int
354
visit_reachable(PyObject *op, PyGC_Head *reachable)
355
{
356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
    if (PyObject_IS_GC(op)) {
        PyGC_Head *gc = AS_GC(op);
        const Py_ssize_t gc_refs = gc->gc.gc_refs;

        if (gc_refs == 0) {
            /* This is in move_unreachable's 'young' list, but
             * the traversal hasn't yet gotten to it.  All
             * we need to do is tell move_unreachable that it's
             * reachable.
             */
            gc->gc.gc_refs = 1;
        }
        else if (gc_refs == GC_TENTATIVELY_UNREACHABLE) {
            /* This had gc_refs = 0 when move_unreachable got
             * to it, but turns out it's reachable after all.
             * Move it back to move_unreachable's 'young' list,
             * and move_unreachable will eventually get to it
             * again.
             */
            gc_list_move(gc, reachable);
            gc->gc.gc_refs = 1;
        }
        /* Else there's nothing to do.
         * If gc_refs > 0, it must be in move_unreachable's 'young'
         * list, and move_unreachable will eventually get to it.
         * If gc_refs == GC_REACHABLE, it's either in some other
         * generation so we don't care about it, or move_unreachable
         * already dealt with it.
         * If gc_refs == GC_UNTRACKED, it must be ignored.
         */
         else {
            assert(gc_refs > 0
                   || gc_refs == GC_REACHABLE
                   || gc_refs == GC_UNTRACKED);
         }
    }
    return 0;
393 394
}

395 396 397 398 399 400 401
/* Move the unreachable objects from young to unreachable.  After this,
 * all objects in young have gc_refs = GC_REACHABLE, and all objects in
 * unreachable have gc_refs = GC_TENTATIVELY_UNREACHABLE.  All tracked
 * gc objects not in young or unreachable still have gc_refs = GC_REACHABLE.
 * All objects in young after this are directly or indirectly reachable
 * from outside the original young; and all objects in unreachable are
 * not.
402
 */
403
static void
404
move_unreachable(PyGC_Head *young, PyGC_Head *unreachable)
405
{
406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
    PyGC_Head *gc = young->gc.gc_next;

    /* Invariants:  all objects "to the left" of us in young have gc_refs
     * = GC_REACHABLE, and are indeed reachable (directly or indirectly)
     * from outside the young list as it was at entry.  All other objects
     * from the original young "to the left" of us are in unreachable now,
     * and have gc_refs = GC_TENTATIVELY_UNREACHABLE.  All objects to the
     * left of us in 'young' now have been scanned, and no objects here
     * or to the right have been scanned yet.
     */

    while (gc != young) {
        PyGC_Head *next;

        if (gc->gc.gc_refs) {
            /* gc is definitely reachable from outside the
             * original 'young'.  Mark it as such, and traverse
             * its pointers to find any other objects that may
             * be directly reachable from it.  Note that the
             * call to tp_traverse may append objects to young,
             * so we have to wait until it returns to determine
             * the next object to visit.
             */
            PyObject *op = FROM_GC(gc);
            traverseproc traverse = Py_TYPE(op)->tp_traverse;
            assert(gc->gc.gc_refs > 0);
            gc->gc.gc_refs = GC_REACHABLE;
            (void) traverse(op,
                            (visitproc)visit_reachable,
                            (void *)young);
            next = gc->gc.gc_next;
            if (PyTuple_CheckExact(op)) {
                _PyTuple_MaybeUntrack(op);
            }
            else if (PyDict_CheckExact(op)) {
                _PyDict_MaybeUntrack(op);
            }
        }
        else {
            /* This *may* be unreachable.  To make progress,
             * assume it is.  gc isn't directly reachable from
             * any object we've already traversed, but may be
             * reachable from an object we haven't gotten to yet.
             * visit_reachable will eventually move gc back into
             * young if that's so, and we'll see it again.
             */
            next = gc->gc.gc_next;
            gc_list_move(gc, unreachable);
            gc->gc.gc_refs = GC_TENTATIVELY_UNREACHABLE;
        }
        gc = next;
    }
458 459
}

460
/* Return true if object has a finalization method. */
461 462
static int
has_finalizer(PyObject *op)
463
{
464 465 466 467
    if (PyGen_CheckExact(op))
        return PyGen_NeedsFinalizing((PyGenObject *)op);
    else
        return op->ob_type->tp_del != NULL;
468 469
}

470 471 472
/* Move the objects in unreachable with __del__ methods into `finalizers`.
 * Objects moved into `finalizers` have gc_refs set to GC_REACHABLE; the
 * objects remaining in unreachable are left at GC_TENTATIVELY_UNREACHABLE.
Jeremy Hylton's avatar
Jeremy Hylton committed
473
 */
474
static void
475
move_finalizers(PyGC_Head *unreachable, PyGC_Head *finalizers)
476
{
477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493
    PyGC_Head *gc;
    PyGC_Head *next;

    /* March over unreachable.  Move objects with finalizers into
     * `finalizers`.
     */
    for (gc = unreachable->gc.gc_next; gc != unreachable; gc = next) {
        PyObject *op = FROM_GC(gc);

        assert(IS_TENTATIVELY_UNREACHABLE(op));
        next = gc->gc.gc_next;

        if (has_finalizer(op)) {
            gc_list_move(gc, finalizers);
            gc->gc.gc_refs = GC_REACHABLE;
        }
    }
494 495 496 497 498 499
}

/* A traversal callback for move_finalizer_reachable. */
static int
visit_move(PyObject *op, PyGC_Head *tolist)
{
500 501 502 503 504 505 506 507
    if (PyObject_IS_GC(op)) {
        if (IS_TENTATIVELY_UNREACHABLE(op)) {
            PyGC_Head *gc = AS_GC(op);
            gc_list_move(gc, tolist);
            gc->gc.gc_refs = GC_REACHABLE;
        }
    }
    return 0;
508 509
}

510
/* Move objects that are reachable from finalizers, from the unreachable set
511
 * into finalizers set.
512
 */
513
static void
514
move_finalizer_reachable(PyGC_Head *finalizers)
515
{
516 517 518 519 520 521 522 523 524
    traverseproc traverse;
    PyGC_Head *gc = finalizers->gc.gc_next;
    for (; gc != finalizers; gc = gc->gc.gc_next) {
        /* Note that the finalizers list may grow during this. */
        traverse = Py_TYPE(FROM_GC(gc))->tp_traverse;
        (void) traverse(FROM_GC(gc),
                        (visitproc)visit_move,
                        (void *)finalizers);
    }
525 526
}

527 528 529 530 531 532 533 534 535 536
/* Clear all weakrefs to unreachable objects, and if such a weakref has a
 * callback, invoke it if necessary.  Note that it's possible for such
 * weakrefs to be outside the unreachable set -- indeed, those are precisely
 * the weakrefs whose callbacks must be invoked.  See gc_weakref.txt for
 * overview & some details.  Some weakrefs with callbacks may be reclaimed
 * directly by this routine; the number reclaimed is the return value.  Other
 * weakrefs with callbacks may be moved into the `old` generation.  Objects
 * moved into `old` have gc_refs set to GC_REACHABLE; the objects remaining in
 * unreachable are left at GC_TENTATIVELY_UNREACHABLE.  When this returns,
 * no object in `unreachable` is weakly referenced anymore.
537
 */
538 539
static int
handle_weakrefs(PyGC_Head *unreachable, PyGC_Head *old)
540
{
541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677
    PyGC_Head *gc;
    PyObject *op;               /* generally FROM_GC(gc) */
    PyWeakReference *wr;        /* generally a cast of op */
    PyGC_Head wrcb_to_call;     /* weakrefs with callbacks to call */
    PyGC_Head *next;
    int num_freed = 0;

    gc_list_init(&wrcb_to_call);

    /* Clear all weakrefs to the objects in unreachable.  If such a weakref
     * also has a callback, move it into `wrcb_to_call` if the callback
     * needs to be invoked.  Note that we cannot invoke any callbacks until
     * all weakrefs to unreachable objects are cleared, lest the callback
     * resurrect an unreachable object via a still-active weakref.  We
     * make another pass over wrcb_to_call, invoking callbacks, after this
     * pass completes.
     */
    for (gc = unreachable->gc.gc_next; gc != unreachable; gc = next) {
        PyWeakReference **wrlist;

        op = FROM_GC(gc);
        assert(IS_TENTATIVELY_UNREACHABLE(op));
        next = gc->gc.gc_next;

        if (! PyType_SUPPORTS_WEAKREFS(Py_TYPE(op)))
            continue;

        /* It supports weakrefs.  Does it have any? */
        wrlist = (PyWeakReference **)
                                PyObject_GET_WEAKREFS_LISTPTR(op);

        /* `op` may have some weakrefs.  March over the list, clear
         * all the weakrefs, and move the weakrefs with callbacks
         * that must be called into wrcb_to_call.
         */
        for (wr = *wrlist; wr != NULL; wr = *wrlist) {
            PyGC_Head *wrasgc;                  /* AS_GC(wr) */

            /* _PyWeakref_ClearRef clears the weakref but leaves
             * the callback pointer intact.  Obscure:  it also
             * changes *wrlist.
             */
            assert(wr->wr_object == op);
            _PyWeakref_ClearRef(wr);
            assert(wr->wr_object == Py_None);
            if (wr->wr_callback == NULL)
                continue;                       /* no callback */

    /* Headache time.  `op` is going away, and is weakly referenced by
     * `wr`, which has a callback.  Should the callback be invoked?  If wr
     * is also trash, no:
     *
     * 1. There's no need to call it.  The object and the weakref are
     *    both going away, so it's legitimate to pretend the weakref is
     *    going away first.  The user has to ensure a weakref outlives its
     *    referent if they want a guarantee that the wr callback will get
     *    invoked.
     *
     * 2. It may be catastrophic to call it.  If the callback is also in
     *    cyclic trash (CT), then although the CT is unreachable from
     *    outside the current generation, CT may be reachable from the
     *    callback.  Then the callback could resurrect insane objects.
     *
     * Since the callback is never needed and may be unsafe in this case,
     * wr is simply left in the unreachable set.  Note that because we
     * already called _PyWeakref_ClearRef(wr), its callback will never
     * trigger.
     *
     * OTOH, if wr isn't part of CT, we should invoke the callback:  the
     * weakref outlived the trash.  Note that since wr isn't CT in this
     * case, its callback can't be CT either -- wr acted as an external
     * root to this generation, and therefore its callback did too.  So
     * nothing in CT is reachable from the callback either, so it's hard
     * to imagine how calling it later could create a problem for us.  wr
     * is moved to wrcb_to_call in this case.
     */
            if (IS_TENTATIVELY_UNREACHABLE(wr))
                continue;
            assert(IS_REACHABLE(wr));

            /* Create a new reference so that wr can't go away
             * before we can process it again.
             */
            Py_INCREF(wr);

            /* Move wr to wrcb_to_call, for the next pass. */
            wrasgc = AS_GC(wr);
            assert(wrasgc != next); /* wrasgc is reachable, but
                                       next isn't, so they can't
                                       be the same */
            gc_list_move(wrasgc, &wrcb_to_call);
        }
    }

    /* Invoke the callbacks we decided to honor.  It's safe to invoke them
     * because they can't reference unreachable objects.
     */
    while (! gc_list_is_empty(&wrcb_to_call)) {
        PyObject *temp;
        PyObject *callback;

        gc = wrcb_to_call.gc.gc_next;
        op = FROM_GC(gc);
        assert(IS_REACHABLE(op));
        assert(PyWeakref_Check(op));
        wr = (PyWeakReference *)op;
        callback = wr->wr_callback;
        assert(callback != NULL);

        /* copy-paste of weakrefobject.c's handle_callback() */
        temp = PyObject_CallFunctionObjArgs(callback, wr, NULL);
        if (temp == NULL)
            PyErr_WriteUnraisable(callback);
        else
            Py_DECREF(temp);

        /* Give up the reference we created in the first pass.  When
         * op's refcount hits 0 (which it may or may not do right now),
         * op's tp_dealloc will decref op->wr_callback too.  Note
         * that the refcount probably will hit 0 now, and because this
         * weakref was reachable to begin with, gc didn't already
         * add it to its count of freed objects.  Example:  a reachable
         * weak value dict maps some key to this reachable weakref.
         * The callback removes this key->weakref mapping from the
         * dict, leaving no other references to the weakref (excepting
         * ours).
         */
        Py_DECREF(op);
        if (wrcb_to_call.gc.gc_next == gc) {
            /* object is still alive -- move it */
            gc_list_move(gc, old);
        }
        else
            ++num_freed;
    }

    return num_freed;
678 679
}

680
static void
681
debug_cycle(char *msg, PyObject *op)
682
{
683 684
    PySys_FormatStderr("gc: %s <%s %p>\n",
                       msg, Py_TYPE(op)->tp_name, op);
685 686
}

687 688
/* Handle uncollectable garbage (cycles with finalizers, and stuff reachable
 * only from such cycles).
689 690 691 692
 * If DEBUG_SAVEALL, all objects in finalizers are appended to the module
 * garbage list (a Python list), else only the objects in finalizers with
 * __del__ methods are appended to garbage.  All objects in finalizers are
 * merged into the old list regardless.
693 694
 * Returns 0 if all OK, <0 on error (out of memory to grow the garbage list).
 * The finalizers list is made empty on a successful return.
695
 */
696
static int
697
handle_finalizers(PyGC_Head *finalizers, PyGC_Head *old)
698
{
699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716
    PyGC_Head *gc = finalizers->gc.gc_next;

    if (garbage == NULL) {
        garbage = PyList_New(0);
        if (garbage == NULL)
            Py_FatalError("gc couldn't create gc.garbage list");
    }
    for (; gc != finalizers; gc = gc->gc.gc_next) {
        PyObject *op = FROM_GC(gc);

        if ((debug & DEBUG_SAVEALL) || has_finalizer(op)) {
            if (PyList_Append(garbage, op) < 0)
                return -1;
        }
    }

    gc_list_merge(finalizers, old);
    return 0;
717 718
}

719
/* Break reference cycles by clearing the containers involved.  This is
720
 * tricky business as the lists can be changing and we don't know which
721 722
 * objects may be freed.  It is possible I screwed something up here.
 */
723
static void
Jeremy Hylton's avatar
Jeremy Hylton committed
724
delete_garbage(PyGC_Head *collectable, PyGC_Head *old)
725
{
726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748
    inquiry clear;

    while (!gc_list_is_empty(collectable)) {
        PyGC_Head *gc = collectable->gc.gc_next;
        PyObject *op = FROM_GC(gc);

        assert(IS_TENTATIVELY_UNREACHABLE(op));
        if (debug & DEBUG_SAVEALL) {
            PyList_Append(garbage, op);
        }
        else {
            if ((clear = Py_TYPE(op)->tp_clear) != NULL) {
                Py_INCREF(op);
                clear(op);
                Py_DECREF(op);
            }
        }
        if (collectable->gc.gc_next == gc) {
            /* object is still alive, move it, it may die later */
            gc_list_move(gc, old);
            gc->gc.gc_refs = GC_REACHABLE;
        }
    }
749 750
}

Christian Heimes's avatar
Christian Heimes committed
751 752 753 754 755 756 757 758
/* Clear all free lists
 * All free lists are cleared during the collection of the highest generation.
 * Allocated items in the free list may keep a pymalloc arena occupied.
 * Clearing the free lists may give back memory to the OS earlier.
 */
static void
clear_freelists(void)
{
759 760 761 762 763 764
    (void)PyMethod_ClearFreeList();
    (void)PyFrame_ClearFreeList();
    (void)PyCFunction_ClearFreeList();
    (void)PyTuple_ClearFreeList();
    (void)PyUnicode_ClearFreeList();
    (void)PyFloat_ClearFreeList();
Christian Heimes's avatar
Christian Heimes committed
765 766
}

767 768 769
static double
get_time(void)
{
770 771
    double result = 0;
    if (tmod != NULL) {
772 773 774
        _Py_identifier(time);

        PyObject *f = _PyObject_CallMethodId(tmod, &PyId_time, NULL);
775 776 777 778 779 780 781 782 783 784
        if (f == NULL) {
            PyErr_Clear();
        }
        else {
            if (PyFloat_Check(f))
                result = PyFloat_AsDouble(f);
            Py_DECREF(f);
        }
    }
    return result;
785 786
}

787 788
/* This is the main function.  Read this to understand how the
 * collection process works. */
789
static Py_ssize_t
790
collect(int generation)
791
{
792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944
    int i;
    Py_ssize_t m = 0; /* # objects collected */
    Py_ssize_t n = 0; /* # unreachable objects that couldn't be collected */
    PyGC_Head *young; /* the generation we are examining */
    PyGC_Head *old; /* next older generation */
    PyGC_Head unreachable; /* non-problematic unreachable trash */
    PyGC_Head finalizers;  /* objects with, & reachable from, __del__ */
    PyGC_Head *gc;
    double t1 = 0.0;

    if (delstr == NULL) {
        delstr = PyUnicode_InternFromString("__del__");
        if (delstr == NULL)
            Py_FatalError("gc couldn't allocate \"__del__\"");
    }

    if (debug & DEBUG_STATS) {
        PySys_WriteStderr("gc: collecting generation %d...\n",
                          generation);
        PySys_WriteStderr("gc: objects in each generation:");
        for (i = 0; i < NUM_GENERATIONS; i++)
            PySys_WriteStderr(" %" PY_FORMAT_SIZE_T "d",
                              gc_list_size(GEN_HEAD(i)));
        t1 = get_time();
        PySys_WriteStderr("\n");
    }

    /* update collection and allocation counters */
    if (generation+1 < NUM_GENERATIONS)
        generations[generation+1].count += 1;
    for (i = 0; i <= generation; i++)
        generations[i].count = 0;

    /* merge younger generations with one we are currently collecting */
    for (i = 0; i < generation; i++) {
        gc_list_merge(GEN_HEAD(i), GEN_HEAD(generation));
    }

    /* handy references */
    young = GEN_HEAD(generation);
    if (generation < NUM_GENERATIONS-1)
        old = GEN_HEAD(generation+1);
    else
        old = young;

    /* Using ob_refcnt and gc_refs, calculate which objects in the
     * container set are reachable from outside the set (i.e., have a
     * refcount greater than 0 when all the references within the
     * set are taken into account).
     */
    update_refs(young);
    subtract_refs(young);

    /* Leave everything reachable from outside young in young, and move
     * everything else (in young) to unreachable.
     * NOTE:  This used to move the reachable objects into a reachable
     * set instead.  But most things usually turn out to be reachable,
     * so it's more efficient to move the unreachable things.
     */
    gc_list_init(&unreachable);
    move_unreachable(young, &unreachable);

    /* Move reachable objects to next generation. */
    if (young != old) {
        if (generation == NUM_GENERATIONS - 2) {
            long_lived_pending += gc_list_size(young);
        }
        gc_list_merge(young, old);
    }
    else {
        long_lived_pending = 0;
        long_lived_total = gc_list_size(young);
    }

    /* All objects in unreachable are trash, but objects reachable from
     * finalizers can't safely be deleted.  Python programmers should take
     * care not to create such things.  For Python, finalizers means
     * instance objects with __del__ methods.  Weakrefs with callbacks
     * can also call arbitrary Python code but they will be dealt with by
     * handle_weakrefs().
     */
    gc_list_init(&finalizers);
    move_finalizers(&unreachable, &finalizers);
    /* finalizers contains the unreachable objects with a finalizer;
     * unreachable objects reachable *from* those are also uncollectable,
     * and we move those into the finalizers list too.
     */
    move_finalizer_reachable(&finalizers);

    /* Collect statistics on collectable objects found and print
     * debugging information.
     */
    for (gc = unreachable.gc.gc_next; gc != &unreachable;
                    gc = gc->gc.gc_next) {
        m++;
        if (debug & DEBUG_COLLECTABLE) {
            debug_cycle("collectable", FROM_GC(gc));
        }
    }

    /* Clear weakrefs and invoke callbacks as necessary. */
    m += handle_weakrefs(&unreachable, old);

    /* Call tp_clear on objects in the unreachable set.  This will cause
     * the reference cycles to be broken.  It may also cause some objects
     * in finalizers to be freed.
     */
    delete_garbage(&unreachable, old);

    /* Collect statistics on uncollectable objects found and print
     * debugging information. */
    for (gc = finalizers.gc.gc_next;
         gc != &finalizers;
         gc = gc->gc.gc_next) {
        n++;
        if (debug & DEBUG_UNCOLLECTABLE)
            debug_cycle("uncollectable", FROM_GC(gc));
    }
    if (debug & DEBUG_STATS) {
        double t2 = get_time();
        if (m == 0 && n == 0)
            PySys_WriteStderr("gc: done");
        else
            PySys_WriteStderr(
                "gc: done, "
                "%" PY_FORMAT_SIZE_T "d unreachable, "
                "%" PY_FORMAT_SIZE_T "d uncollectable",
                n+m, n);
        if (t1 && t2) {
            PySys_WriteStderr(", %.4fs elapsed", t2-t1);
        }
        PySys_WriteStderr(".\n");
    }

    /* Append instances in the uncollectable set to a Python
     * reachable list of garbage.  The programmer has to deal with
     * this if they insist on creating this type of structure.
     */
    (void)handle_finalizers(&finalizers, old);

    /* Clear free list only during the collection of the highest
     * generation */
    if (generation == NUM_GENERATIONS-1) {
        clear_freelists();
    }

    if (PyErr_Occurred()) {
        if (gc_str == NULL)
            gc_str = PyUnicode_FromString("garbage collection");
        PyErr_WriteUnraisable(gc_str);
        Py_FatalError("unexpected exception during garbage collection");
    }
    return n+m;
945 946
}

947
static Py_ssize_t
948 949
collect_generations(void)
{
950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969
    int i;
    Py_ssize_t n = 0;

    /* Find the oldest generation (highest numbered) where the count
     * exceeds the threshold.  Objects in the that generation and
     * generations younger than it will be collected. */
    for (i = NUM_GENERATIONS-1; i >= 0; i--) {
        if (generations[i].count > generations[i].threshold) {
            /* Avoid quadratic performance degradation in number
               of tracked objects. See comments at the beginning
               of this file, and issue #4074.
            */
            if (i == NUM_GENERATIONS - 1
                && long_lived_pending < long_lived_total / 4)
                continue;
            n = collect(i);
            break;
        }
    }
    return n;
970 971
}

972
PyDoc_STRVAR(gc_enable__doc__,
973 974
"enable() -> None\n"
"\n"
975
"Enable automatic garbage collection.\n");
976 977

static PyObject *
978
gc_enable(PyObject *self, PyObject *noargs)
979
{
980 981 982
    enabled = 1;
    Py_INCREF(Py_None);
    return Py_None;
983 984
}

985
PyDoc_STRVAR(gc_disable__doc__,
986 987
"disable() -> None\n"
"\n"
988
"Disable automatic garbage collection.\n");
989 990

static PyObject *
991
gc_disable(PyObject *self, PyObject *noargs)
992
{
993 994 995
    enabled = 0;
    Py_INCREF(Py_None);
    return Py_None;
996 997
}

998
PyDoc_STRVAR(gc_isenabled__doc__,
999 1000
"isenabled() -> status\n"
"\n"
1001
"Returns true if automatic garbage collection is enabled.\n");
1002 1003

static PyObject *
1004
gc_isenabled(PyObject *self, PyObject *noargs)
1005
{
1006
    return PyBool_FromLong((long)enabled);
1007
}
1008

1009
PyDoc_STRVAR(gc_collect__doc__,
1010
"collect([generation]) -> n\n"
1011
"\n"
1012 1013 1014 1015
"With no arguments, run a full collection.  The optional argument\n"
"may be an integer specifying which generation to collect.  A ValueError\n"
"is raised if the generation number is invalid.\n\n"
"The number of unreachable objects is returned.\n");
1016 1017

static PyObject *
1018
gc_collect(PyObject *self, PyObject *args, PyObject *kws)
1019
{
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040
    static char *keywords[] = {"generation", NULL};
    int genarg = NUM_GENERATIONS - 1;
    Py_ssize_t n;

    if (!PyArg_ParseTupleAndKeywords(args, kws, "|i", keywords, &genarg))
        return NULL;

    else if (genarg < 0 || genarg >= NUM_GENERATIONS) {
        PyErr_SetString(PyExc_ValueError, "invalid generation");
        return NULL;
    }

    if (collecting)
        n = 0; /* already collecting, don't do anything */
    else {
        collecting = 1;
        n = collect(genarg);
        collecting = 0;
    }

    return PyLong_FromSsize_t(n);
1041 1042
}

1043
PyDoc_STRVAR(gc_set_debug__doc__,
1044 1045 1046 1047 1048 1049 1050 1051 1052 1053
"set_debug(flags) -> None\n"
"\n"
"Set the garbage collection debugging flags. Debugging information is\n"
"written to sys.stderr.\n"
"\n"
"flags is an integer and can have the following bits turned on:\n"
"\n"
"  DEBUG_STATS - Print statistics during collection.\n"
"  DEBUG_COLLECTABLE - Print collectable objects found.\n"
"  DEBUG_UNCOLLECTABLE - Print unreachable but uncollectable objects found.\n"
1054
"  DEBUG_SAVEALL - Save objects to gc.garbage rather than freeing them.\n"
1055
"  DEBUG_LEAK - Debug leaking programs (everything but STATS).\n");
1056 1057

static PyObject *
1058
gc_set_debug(PyObject *self, PyObject *args)
1059
{
1060 1061
    if (!PyArg_ParseTuple(args, "i:set_debug", &debug))
        return NULL;
1062

1063 1064
    Py_INCREF(Py_None);
    return Py_None;
1065 1066
}

1067
PyDoc_STRVAR(gc_get_debug__doc__,
1068 1069
"get_debug() -> flags\n"
"\n"
1070
"Get the garbage collection debugging flags.\n");
1071 1072

static PyObject *
1073
gc_get_debug(PyObject *self, PyObject *noargs)
1074
{
1075
    return Py_BuildValue("i", debug);
1076 1077
}

1078
PyDoc_STRVAR(gc_set_thresh__doc__,
1079
"set_threshold(threshold0, [threshold1, threshold2]) -> None\n"
1080 1081
"\n"
"Sets the collection thresholds.  Setting threshold0 to zero disables\n"
1082
"collection.\n");
1083 1084

static PyObject *
1085
gc_set_thresh(PyObject *self, PyObject *args)
1086
{
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099
    int i;
    if (!PyArg_ParseTuple(args, "i|ii:set_threshold",
                          &generations[0].threshold,
                          &generations[1].threshold,
                          &generations[2].threshold))
        return NULL;
    for (i = 2; i < NUM_GENERATIONS; i++) {
        /* generations higher than 2 get the same threshold */
        generations[i].threshold = generations[2].threshold;
    }

    Py_INCREF(Py_None);
    return Py_None;
1100 1101
}

1102
PyDoc_STRVAR(gc_get_thresh__doc__,
1103 1104
"get_threshold() -> (threshold0, threshold1, threshold2)\n"
"\n"
1105
"Return the current collection thresholds\n");
1106 1107

static PyObject *
1108
gc_get_thresh(PyObject *self, PyObject *noargs)
1109
{
1110 1111 1112 1113
    return Py_BuildValue("(iii)",
                         generations[0].threshold,
                         generations[1].threshold,
                         generations[2].threshold);
1114 1115
}

1116 1117 1118 1119 1120 1121 1122 1123
PyDoc_STRVAR(gc_get_count__doc__,
"get_count() -> (count0, count1, count2)\n"
"\n"
"Return the current collection counts\n");

static PyObject *
gc_get_count(PyObject *self, PyObject *noargs)
{
1124 1125 1126 1127
    return Py_BuildValue("(iii)",
                         generations[0].count,
                         generations[1].count,
                         generations[2].count);
1128 1129
}

1130
static int
1131
referrersvisit(PyObject* obj, PyObject *objs)
1132
{
1133 1134 1135 1136 1137
    Py_ssize_t i;
    for (i = 0; i < PyTuple_GET_SIZE(objs); i++)
        if (PyTuple_GET_ITEM(objs, i) == obj)
            return 1;
    return 0;
1138 1139
}

1140
static int
1141
gc_referrers_for(PyObject *objs, PyGC_Head *list, PyObject *resultlist)
1142
{
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156
    PyGC_Head *gc;
    PyObject *obj;
    traverseproc traverse;
    for (gc = list->gc.gc_next; gc != list; gc = gc->gc.gc_next) {
        obj = FROM_GC(gc);
        traverse = Py_TYPE(obj)->tp_traverse;
        if (obj == objs || obj == resultlist)
            continue;
        if (traverse(obj, (visitproc)referrersvisit, objs)) {
            if (PyList_Append(resultlist, obj) < 0)
                return 0; /* error */
        }
    }
    return 1; /* no error */
1157 1158
}

1159
PyDoc_STRVAR(gc_get_referrers__doc__,
1160
"get_referrers(*objs) -> list\n\
1161
Return the list of objects that directly refer to any of objs.");
1162

1163
static PyObject *
1164
gc_get_referrers(PyObject *self, PyObject *args)
1165
{
1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
    int i;
    PyObject *result = PyList_New(0);
    if (!result) return NULL;

    for (i = 0; i < NUM_GENERATIONS; i++) {
        if (!(gc_referrers_for(args, GEN_HEAD(i), result))) {
            Py_DECREF(result);
            return NULL;
        }
    }
    return result;
1177
}
1178

1179
/* Append obj to list; return true if error (out of memory), false if OK. */
1180
static int
1181
referentsvisit(PyObject *obj, PyObject *list)
1182
{
1183
    return PyList_Append(list, obj) < 0;
1184 1185
}

1186 1187
PyDoc_STRVAR(gc_get_referents__doc__,
"get_referents(*objs) -> list\n\
1188
Return the list of objects that are directly referred to by objs.");
1189 1190

static PyObject *
1191
gc_get_referents(PyObject *self, PyObject *args)
1192
{
1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213
    Py_ssize_t i;
    PyObject *result = PyList_New(0);

    if (result == NULL)
        return NULL;

    for (i = 0; i < PyTuple_GET_SIZE(args); i++) {
        traverseproc traverse;
        PyObject *obj = PyTuple_GET_ITEM(args, i);

        if (! PyObject_IS_GC(obj))
            continue;
        traverse = Py_TYPE(obj)->tp_traverse;
        if (! traverse)
            continue;
        if (traverse(obj, (visitproc)referentsvisit, result)) {
            Py_DECREF(result);
            return NULL;
        }
    }
    return result;
1214 1215
}

1216
PyDoc_STRVAR(gc_get_objects__doc__,
1217 1218 1219
"get_objects() -> [...]\n"
"\n"
"Return a list of objects tracked by the collector (excluding the list\n"
1220
"returned).\n");
1221 1222

static PyObject *
1223
gc_get_objects(PyObject *self, PyObject *noargs)
1224
{
1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237
    int i;
    PyObject* result;

    result = PyList_New(0);
    if (result == NULL)
        return NULL;
    for (i = 0; i < NUM_GENERATIONS; i++) {
        if (append_objects(result, GEN_HEAD(i))) {
            Py_DECREF(result);
            return NULL;
        }
    }
    return result;
1238 1239
}

1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
PyDoc_STRVAR(gc_is_tracked__doc__,
"is_tracked(obj) -> bool\n"
"\n"
"Returns true if the object is tracked by the garbage collector.\n"
"Simple atomic objects will return false.\n"
);

static PyObject *
gc_is_tracked(PyObject *self, PyObject *obj)
{
1250 1251 1252 1253 1254 1255 1256 1257
    PyObject *result;

    if (PyObject_IS_GC(obj) && IS_TRACKED(obj))
        result = Py_True;
    else
        result = Py_False;
    Py_INCREF(result);
    return result;
1258 1259
}

1260

1261
PyDoc_STRVAR(gc__doc__,
1262 1263
"This module provides access to the garbage collector for reference cycles.\n"
"\n"
1264 1265 1266
"enable() -- Enable automatic garbage collection.\n"
"disable() -- Disable automatic garbage collection.\n"
"isenabled() -- Returns true if automatic collection is enabled.\n"
1267
"collect() -- Do a full collection right now.\n"
1268
"get_count() -- Return the current collection counts.\n"
1269 1270 1271 1272
"set_debug() -- Set debugging flags.\n"
"get_debug() -- Get debugging flags.\n"
"set_threshold() -- Set the collection thresholds.\n"
"get_threshold() -- Return the current the collection thresholds.\n"
1273
"get_objects() -- Return a list of all objects tracked by the collector.\n"
1274
"is_tracked() -- Returns true if a given object is tracked.\n"
1275
"get_referrers() -- Return the list of objects that refer to an object.\n"
1276
"get_referents() -- Return the list of objects that an object refers to.\n");
1277 1278

static PyMethodDef GcMethods[] = {
1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295
    {"enable",             gc_enable,     METH_NOARGS,  gc_enable__doc__},
    {"disable",            gc_disable,    METH_NOARGS,  gc_disable__doc__},
    {"isenabled",          gc_isenabled,  METH_NOARGS,  gc_isenabled__doc__},
    {"set_debug",          gc_set_debug,  METH_VARARGS, gc_set_debug__doc__},
    {"get_debug",          gc_get_debug,  METH_NOARGS,  gc_get_debug__doc__},
    {"get_count",          gc_get_count,  METH_NOARGS,  gc_get_count__doc__},
    {"set_threshold",  gc_set_thresh, METH_VARARGS, gc_set_thresh__doc__},
    {"get_threshold",  gc_get_thresh, METH_NOARGS,  gc_get_thresh__doc__},
    {"collect",            (PyCFunction)gc_collect,
        METH_VARARGS | METH_KEYWORDS,           gc_collect__doc__},
    {"get_objects",    gc_get_objects,METH_NOARGS,  gc_get_objects__doc__},
    {"is_tracked",     gc_is_tracked, METH_O,       gc_is_tracked__doc__},
    {"get_referrers",  gc_get_referrers, METH_VARARGS,
        gc_get_referrers__doc__},
    {"get_referents",  gc_get_referents, METH_VARARGS,
        gc_get_referents__doc__},
    {NULL,      NULL}           /* Sentinel */
1296 1297
};

1298
static struct PyModuleDef gcmodule = {
1299
    PyModuleDef_HEAD_INIT,
1300 1301 1302 1303 1304 1305 1306 1307
    "gc",              /* m_name */
    gc__doc__,         /* m_doc */
    -1,                /* m_size */
    GcMethods,         /* m_methods */
    NULL,              /* m_reload */
    NULL,              /* m_traverse */
    NULL,              /* m_clear */
    NULL               /* m_free */
1308 1309
};

1310
PyMODINIT_FUNC
1311
PyInit_gc(void)
1312
{
1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339
    PyObject *m;

    m = PyModule_Create(&gcmodule);

    if (m == NULL)
        return NULL;

    if (garbage == NULL) {
        garbage = PyList_New(0);
        if (garbage == NULL)
            return NULL;
    }
    Py_INCREF(garbage);
    if (PyModule_AddObject(m, "garbage", garbage) < 0)
        return NULL;

    /* Importing can't be done in collect() because collect()
     * can be called via PyGC_Collect() in Py_Finalize().
     * This wouldn't be a problem, except that <initialized> is
     * reset to 0 before calling collect which trips up
     * the import and triggers an assertion.
     */
    if (tmod == NULL) {
        tmod = PyImport_ImportModuleNoBlock("time");
        if (tmod == NULL)
            PyErr_Clear();
    }
1340

1341
#define ADD_INT(NAME) if (PyModule_AddIntConstant(m, #NAME, NAME) < 0) return NULL
1342 1343 1344 1345 1346
    ADD_INT(DEBUG_STATS);
    ADD_INT(DEBUG_COLLECTABLE);
    ADD_INT(DEBUG_UNCOLLECTABLE);
    ADD_INT(DEBUG_SAVEALL);
    ADD_INT(DEBUG_LEAK);
1347
#undef ADD_INT
1348
    return m;
1349 1350
}

1351
/* API to invoke gc.collect() from C */
1352
Py_ssize_t
1353 1354
PyGC_Collect(void)
{
1355
    Py_ssize_t n;
1356

1357 1358 1359 1360 1361 1362 1363
    if (collecting)
        n = 0; /* already collecting, don't do anything */
    else {
        collecting = 1;
        n = collect(NUM_GENERATIONS - 1);
        collecting = 0;
    }
1364

1365
    return n;
1366 1367
}

1368 1369 1370
void
_PyGC_Fini(void)
{
1371 1372
    if (!(debug & DEBUG_SAVEALL)
        && garbage != NULL && PyList_GET_SIZE(garbage) > 0) {
1373 1374
        char *message;
        if (debug & DEBUG_UNCOLLECTABLE)
1375
            message = "gc: %zd uncollectable objects at " \
1376 1377
                "shutdown";
        else
1378
            message = "gc: %zd uncollectable objects at " \
1379 1380 1381 1382
                "shutdown; use gc.set_debug(gc.DEBUG_UNCOLLECTABLE) to list them";
        if (PyErr_WarnFormat(PyExc_ResourceWarning, 0, message,
                             PyList_GET_SIZE(garbage)) < 0)
            PyErr_WriteUnraisable(NULL);
1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
        if (debug & DEBUG_UNCOLLECTABLE) {
            PyObject *repr = NULL, *bytes = NULL;
            repr = PyObject_Repr(garbage);
            if (!repr || !(bytes = PyUnicode_EncodeFSDefault(repr)))
                PyErr_WriteUnraisable(garbage);
            else {
                PySys_WriteStderr(
                    "    %s\n",
                    PyBytes_AS_STRING(bytes)
                    );
            }
            Py_XDECREF(repr);
            Py_XDECREF(bytes);
        }
    }
}

1400
/* for debugging */
1401 1402
void
_PyGC_Dump(PyGC_Head *g)
1403
{
1404
    _PyObject_Dump(FROM_GC(g));
1405 1406 1407 1408 1409
}

/* extension modules might be compiled with GC support so these
   functions must always be available */

1410 1411 1412 1413 1414
#undef PyObject_GC_Track
#undef PyObject_GC_UnTrack
#undef PyObject_GC_Del
#undef _PyObject_GC_Malloc

1415
void
1416
PyObject_GC_Track(void *op)
1417
{
1418
    _PyObject_GC_TRACK(op);
1419 1420
}

1421
/* for binary compatibility with 2.2 */
1422
void
1423 1424 1425 1426 1427 1428 1429
_PyObject_GC_Track(PyObject *op)
{
    PyObject_GC_Track(op);
}

void
PyObject_GC_UnTrack(void *op)
1430
{
1431 1432 1433 1434 1435
    /* Obscure:  the Py_TRASHCAN mechanism requires that we be able to
     * call PyObject_GC_UnTrack twice on an object.
     */
    if (IS_TRACKED(op))
        _PyObject_GC_UNTRACK(op);
1436 1437
}

1438 1439 1440 1441 1442 1443 1444
/* for binary compatibility with 2.2 */
void
_PyObject_GC_UnTrack(PyObject *op)
{
    PyObject_GC_UnTrack(op);
}

1445
PyObject *
1446
_PyObject_GC_Malloc(size_t basicsize)
1447
{
1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468
    PyObject *op;
    PyGC_Head *g;
    if (basicsize > PY_SSIZE_T_MAX - sizeof(PyGC_Head))
        return PyErr_NoMemory();
    g = (PyGC_Head *)PyObject_MALLOC(
        sizeof(PyGC_Head) + basicsize);
    if (g == NULL)
        return PyErr_NoMemory();
    g->gc.gc_refs = GC_UNTRACKED;
    generations[0].count++; /* number of allocated GC objects */
    if (generations[0].count > generations[0].threshold &&
        enabled &&
        generations[0].threshold &&
        !collecting &&
        !PyErr_Occurred()) {
        collecting = 1;
        collect_generations();
        collecting = 0;
    }
    op = FROM_GC(g);
    return op;
1469 1470 1471 1472 1473
}

PyObject *
_PyObject_GC_New(PyTypeObject *tp)
{
1474 1475 1476 1477
    PyObject *op = _PyObject_GC_Malloc(_PyObject_SIZE(tp));
    if (op != NULL)
        op = PyObject_INIT(op, tp);
    return op;
1478 1479 1480
}

PyVarObject *
Martin v. Löwis's avatar
Martin v. Löwis committed
1481
_PyObject_GC_NewVar(PyTypeObject *tp, Py_ssize_t nitems)
1482
{
1483 1484 1485 1486 1487
    const size_t size = _PyObject_VAR_SIZE(tp, nitems);
    PyVarObject *op = (PyVarObject *) _PyObject_GC_Malloc(size);
    if (op != NULL)
        op = PyObject_INIT_VAR(op, tp, nitems);
    return op;
1488 1489 1490
}

PyVarObject *
1491
_PyObject_GC_Resize(PyVarObject *op, Py_ssize_t nitems)
1492
{
1493 1494 1495 1496 1497 1498 1499 1500 1501 1502
    const size_t basicsize = _PyObject_VAR_SIZE(Py_TYPE(op), nitems);
    PyGC_Head *g = AS_GC(op);
    if (basicsize > PY_SSIZE_T_MAX - sizeof(PyGC_Head))
        return (PyVarObject *)PyErr_NoMemory();
    g = (PyGC_Head *)PyObject_REALLOC(g,  sizeof(PyGC_Head) + basicsize);
    if (g == NULL)
        return (PyVarObject *)PyErr_NoMemory();
    op = (PyVarObject *) FROM_GC(g);
    Py_SIZE(op) = nitems;
    return op;
1503 1504 1505
}

void
1506
PyObject_GC_Del(void *op)
1507
{
1508 1509 1510 1511 1512 1513 1514
    PyGC_Head *g = AS_GC(op);
    if (IS_TRACKED(op))
        gc_list_remove(g);
    if (generations[0].count > 0) {
        generations[0].count--;
    }
    PyObject_FREE(g);
1515
}