1 /* Copyright (c) 2008-2023. The SimGrid Team. All rights reserved. */
3 /* This program is free software; you can redistribute it and/or modify it
4 * under the terms of the license (GNU LGPL) which comes with this package. */
6 /** \file compare.cpp Memory snapshotting and comparison */
8 #include "src/mc/mc_config.hpp"
9 #include "src/mc/mc_private.hpp"
10 #include "src/mc/sosp/RemoteProcessMemory.hpp"
11 #include "src/mc/sosp/Snapshot.hpp"
16 XBT_LOG_NEW_DEFAULT_SUBCATEGORY(mc_compare, mc, "Logging specific to mc_compare in mc");
18 namespace simgrid::mc {
20 /*********************************** Heap comparison ***********************************/
21 /***************************************************************************************/
28 HeapLocation() = default;
29 explicit HeapLocation(int block, int fragment = 0) : block_(block), fragment_(fragment) {}
31 bool operator==(HeapLocation const& that) const
33 return block_ == that.block_ && fragment_ == that.fragment_;
35 bool operator<(HeapLocation const& that) const
37 return std::make_pair(block_, fragment_) < std::make_pair(that.block_, that.fragment_);
41 using HeapLocationPair = std::array<HeapLocation, 2>;
42 using HeapLocationPairs = std::set<HeapLocationPair>;
44 class HeapArea : public HeapLocation {
48 explicit HeapArea(int block) : valid_(true) { block_ = block; }
49 HeapArea(int block, int fragment) : valid_(true)
56 class ProcessComparisonState {
58 const std::vector<IgnoredHeapRegion>* to_ignore = nullptr;
59 std::vector<HeapArea> equals_to;
60 std::vector<Type*> types;
61 std::size_t heapsize = 0;
63 void initHeapInformation(const s_xbt_mheap_t* heap, const std::vector<IgnoredHeapRegion>& i);
66 class StateComparator {
68 s_xbt_mheap_t std_heap_copy;
69 std::size_t heaplimit;
70 std::array<ProcessComparisonState, 2> processStates;
72 std::unordered_set<std::pair<const void*, const void*>, simgrid::xbt::hash<std::pair<const void*, const void*>>>
77 compared_pointers.clear();
80 int initHeapInformation(RemoteProcessMemory& appli, const s_xbt_mheap_t* heap1, const s_xbt_mheap_t* heap2,
81 const std::vector<IgnoredHeapRegion>& i1, const std::vector<IgnoredHeapRegion>& i2);
83 template <int rank> HeapArea& equals_to_(std::size_t i, std::size_t j)
85 return processStates[rank - 1].equals_to[MAX_FRAGMENT_PER_BLOCK * i + j];
87 template <int rank> Type*& types_(std::size_t i, std::size_t j)
89 return processStates[rank - 1].types[MAX_FRAGMENT_PER_BLOCK * i + j];
92 template <int rank> HeapArea const& equals_to_(std::size_t i, std::size_t j) const
94 return processStates[rank - 1].equals_to[MAX_FRAGMENT_PER_BLOCK * i + j];
96 template <int rank> Type* const& types_(std::size_t i, std::size_t j) const
98 return processStates[rank - 1].types[MAX_FRAGMENT_PER_BLOCK * i + j];
101 /** Check whether two blocks are known to be matching
103 * @param b1 Block of state 1
104 * @param b2 Block of state 2
105 * @return if the blocks are known to be matching
107 bool blocksEqual(int b1, int b2) const
109 return this->equals_to_<1>(b1, 0).block_ == b2 && this->equals_to_<2>(b2, 0).block_ == b1;
112 /** Check whether two fragments are known to be matching
114 * @param b1 Block of state 1
115 * @param f1 Fragment of state 1
116 * @param b2 Block of state 2
117 * @param f2 Fragment of state 2
118 * @return if the fragments are known to be matching
120 int fragmentsEqual(int b1, int f1, int b2, int f2) const
122 return this->equals_to_<1>(b1, f1).block_ == b2 && this->equals_to_<1>(b1, f1).fragment_ == f2 &&
123 this->equals_to_<2>(b2, f2).block_ == b1 && this->equals_to_<2>(b2, f2).fragment_ == f1;
126 void match_equals(const HeapLocationPairs* list);
129 } // namespace simgrid::mc
131 /************************************************************************************/
133 static ssize_t heap_comparison_ignore_size(const std::vector<simgrid::mc::IgnoredHeapRegion>* ignore_list,
136 auto pos = std::lower_bound(ignore_list->begin(), ignore_list->end(), address,
137 [](auto const& reg, auto const* addr) { return reg.address < addr; });
138 return (pos != ignore_list->end() && pos->address == address) ? pos->size : -1;
141 static bool is_stack(const simgrid::mc::RemoteProcessMemory& process, const void* address)
143 auto const& stack_areas = process.stack_areas();
144 return std::any_of(stack_areas.begin(), stack_areas.end(),
145 [address](auto const& stack) { return stack.address == address; });
148 // TODO, this should depend on the snapshot?
149 static bool is_block_stack(const simgrid::mc::RemoteProcessMemory& process, int block)
151 auto const& stack_areas = process.stack_areas();
152 return std::any_of(stack_areas.begin(), stack_areas.end(),
153 [block](auto const& stack) { return stack.block == block; });
156 namespace simgrid::mc {
158 void StateComparator::match_equals(const HeapLocationPairs* list)
160 for (auto const& pair : *list) {
161 if (pair[0].fragment_ != -1) {
162 this->equals_to_<1>(pair[0].block_, pair[0].fragment_) = HeapArea(pair[1].block_, pair[1].fragment_);
163 this->equals_to_<2>(pair[1].block_, pair[1].fragment_) = HeapArea(pair[0].block_, pair[0].fragment_);
165 this->equals_to_<1>(pair[0].block_, 0) = HeapArea(pair[1].block_, pair[1].fragment_);
166 this->equals_to_<2>(pair[1].block_, 0) = HeapArea(pair[0].block_, pair[0].fragment_);
171 void ProcessComparisonState::initHeapInformation(const s_xbt_mheap_t* heap, const std::vector<IgnoredHeapRegion>& i)
173 auto heaplimit = heap->heaplimit;
174 this->heapsize = heap->heapsize;
175 this->to_ignore = &i;
176 this->equals_to.assign(heaplimit * MAX_FRAGMENT_PER_BLOCK, HeapArea());
177 this->types.assign(heaplimit * MAX_FRAGMENT_PER_BLOCK, nullptr);
180 int StateComparator::initHeapInformation(simgrid::mc::RemoteProcessMemory& memory, const s_xbt_mheap_t* heap1,
181 const s_xbt_mheap_t* heap2, const std::vector<IgnoredHeapRegion>& i1,
182 const std::vector<IgnoredHeapRegion>& i2)
184 if ((heap1->heaplimit != heap2->heaplimit) || (heap1->heapsize != heap2->heapsize))
186 this->heaplimit = heap1->heaplimit;
187 this->std_heap_copy = *memory.get_heap();
188 this->processStates[0].initHeapInformation(heap1, i1);
189 this->processStates[1].initHeapInformation(heap2, i2);
193 // TODO, have a robust way to find it in O(1)
194 static inline Region* MC_get_heap_region(const Snapshot& snapshot)
196 for (auto const& region : snapshot.snapshot_regions_)
197 if (region->region_type() == RegionType::Heap)
199 xbt_die("No heap region");
202 static bool heap_area_differ(const RemoteProcessMemory& process, StateComparator& state, const void* area1,
203 const void* area2, const Snapshot& snapshot1, const Snapshot& snapshot2,
204 HeapLocationPairs* previous, Type* type, int pointer_level);
206 /* Compares the content of each heap fragment between the two states, at the bit level.
208 * This operation is costly (about 5 seconds per snapshots' pair to compare on a small program),
209 * but hard to optimize because our algorithm is too hackish.
211 * Going at bit level can trigger syntaxtic differences on states that are semantically equivalent.
213 * Padding bytes constitute the first source of such syntaxtic difference: Any malloced memory contains spaces that
214 * are not used to enforce the memory alignment constraints of the CPU. So, cruft of irrelevant changes could get
215 * added on these bits. But this case is handled properly, as any memory block is zeroed by mmalloc before being handled
216 * back, not only for calloc but also for malloc. So the memory interstices due to padding bytes are properly zeroed.
218 * Another source of such change comes from the order of mallocs, that may well change from one execution path to
219 * another. This will change the malloc fragment in which the data is stored and the pointer values (syntaxtic
220 * difference) while the semantic of the state remains the same.
222 * To fix this, this code relies on a hugly hack. When we see a difference during the bit-level comparison,
223 * we first check if it could be explained by a pointer-to-block difference. Ie, if when interpreting the memory
224 * area containing that difference as a pointer, I get the pointer to a valid fragment in the heap (in both snapshots).
226 * This is why we cannot pre-compute a bit-level hash of the heap content: we discover the pointers to other memory
227 * fragment when a difference is found during the bit-level exploration. Fixing this would require to save typing
228 * information about the memory fragments, which is something that could be done with https://github.com/tudasc/TypeART
229 * This would give us all pointers in the mallocated memory, allowing the graph traversal needed to precompute the hash.
231 * Using a hash without paying attention to malloc fragment reordering would lead to false negatives:
232 * semantically equivalent states would be detected as [syntaxically] different. It's of no importance for the
233 * state-equality reduction (we would re-explore semantically equivalent states), but it would endanger the soundness
234 * of the liveness model-checker, as state-equality is used to detect the loops that constitute the accepting states of
235 * the verified property. So we could miss counter-examples to the verified property. Not good. Not good at all.
237 static bool mmalloc_heap_differ(const RemoteProcessMemory& process, StateComparator& state, const Snapshot& snapshot1,
238 const Snapshot& snapshot2)
240 /* Check busy blocks */
243 malloc_info heapinfo_temp1;
244 malloc_info heapinfo_temp2;
245 malloc_info heapinfo_temp2b;
247 const Region* heap_region1 = MC_get_heap_region(snapshot1);
248 const Region* heap_region2 = MC_get_heap_region(snapshot2);
250 // This is the address of std_heap->heapinfo in the application process:
251 uint64_t heapinfo_address = process.heap_address.address() + offsetof(s_xbt_mheap_t, heapinfo);
253 // This is in snapshot do not use them directly:
254 const malloc_info* heapinfos1 = snapshot1.read(remote<malloc_info*>(heapinfo_address));
255 const malloc_info* heapinfos2 = snapshot2.read(remote<malloc_info*>(heapinfo_address));
257 while (i1 < state.heaplimit) {
258 const auto* heapinfo1 =
259 static_cast<malloc_info*>(heap_region1->read(&heapinfo_temp1, &heapinfos1[i1], sizeof(malloc_info)));
260 const auto* heapinfo2 =
261 static_cast<malloc_info*>(heap_region2->read(&heapinfo_temp2, &heapinfos2[i1], sizeof(malloc_info)));
263 if (heapinfo1->type == MMALLOC_TYPE_FREE || heapinfo1->type == MMALLOC_TYPE_HEAPINFO) { /* Free block */
268 xbt_assert(heapinfo1->type >= 0, "Unknown mmalloc block type: %d", heapinfo1->type);
270 void* addr_block1 = (ADDR2UINT(i1) - 1) * BLOCKSIZE + (char*)state.std_heap_copy.heapbase;
272 if (heapinfo1->type == MMALLOC_TYPE_UNFRAGMENTED) { /* Large block */
273 if (is_stack(process, addr_block1)) {
274 for (size_t k = 0; k < heapinfo1->busy_block.size; k++)
275 state.equals_to_<1>(i1 + k, 0) = HeapArea(i1, -1);
276 for (size_t k = 0; k < heapinfo2->busy_block.size; k++)
277 state.equals_to_<2>(i1 + k, 0) = HeapArea(i1, -1);
278 i1 += heapinfo1->busy_block.size;
282 if (state.equals_to_<1>(i1, 0).valid_) {
290 /* Try first to associate to same block in the other heap */
291 if (heapinfo2->type == heapinfo1->type && state.equals_to_<2>(i1, 0).valid_ == 0) {
292 const void* addr_block2 = (ADDR2UINT(i1) - 1) * BLOCKSIZE + (char*)state.std_heap_copy.heapbase;
293 if (not heap_area_differ(process, state, addr_block1, addr_block2, snapshot1, snapshot2, nullptr, nullptr, 0)) {
294 for (size_t k = 1; k < heapinfo2->busy_block.size; k++)
295 state.equals_to_<2>(i1 + k, 0) = HeapArea(i1, -1);
296 for (size_t k = 1; k < heapinfo1->busy_block.size; k++)
297 state.equals_to_<1>(i1 + k, 0) = HeapArea(i1, -1);
299 i1 += heapinfo1->busy_block.size;
303 while (i2 < state.heaplimit && not equal) {
304 const void* addr_block2 = (ADDR2UINT(i2) - 1) * BLOCKSIZE + (char*)state.std_heap_copy.heapbase;
311 const auto* heapinfo2b =
312 static_cast<malloc_info*>(heap_region2->read(&heapinfo_temp2b, &heapinfos2[i2], sizeof(malloc_info)));
314 if (heapinfo2b->type != MMALLOC_TYPE_UNFRAGMENTED) {
319 if (state.equals_to_<2>(i2, 0).valid_) {
324 if (not heap_area_differ(process, state, addr_block1, addr_block2, snapshot1, snapshot2, nullptr, nullptr, 0)) {
325 for (size_t k = 1; k < heapinfo2b->busy_block.size; k++)
326 state.equals_to_<2>(i2 + k, 0) = HeapArea(i1, -1);
327 for (size_t k = 1; k < heapinfo1->busy_block.size; k++)
328 state.equals_to_<1>(i1 + k, 0) = HeapArea(i2, -1);
330 i1 += heapinfo1->busy_block.size;
336 XBT_DEBUG("Block %zu not found (size_used = %zu, addr = %p)", i1, heapinfo1->busy_block.busy_size, addr_block1);
339 } else { /* Fragmented block */
340 for (size_t j1 = 0; j1 < (size_t)(BLOCKSIZE >> heapinfo1->type); j1++) {
341 if (heapinfo1->busy_frag.frag_size[j1] == -1) /* Free fragment_ */
344 if (state.equals_to_<1>(i1, j1).valid_)
347 void* addr_frag1 = (char*)addr_block1 + (j1 << heapinfo1->type);
352 /* Try first to associate to same fragment_ in the other heap */
353 if (heapinfo2->type == heapinfo1->type && not state.equals_to_<2>(i1, j1).valid_) {
354 const void* addr_block2 = (ADDR2UINT(i1) - 1) * BLOCKSIZE + (char*)state.std_heap_copy.heapbase;
355 const void* addr_frag2 = (const char*)addr_block2 + (j1 << heapinfo2->type);
356 if (not heap_area_differ(process, state, addr_frag1, addr_frag2, snapshot1, snapshot2, nullptr, nullptr, 0))
360 while (i2 < state.heaplimit && not equal) {
361 const auto* heapinfo2b =
362 static_cast<malloc_info*>(heap_region2->read(&heapinfo_temp2b, &heapinfos2[i2], sizeof(malloc_info)));
364 if (heapinfo2b->type == MMALLOC_TYPE_FREE || heapinfo2b->type == MMALLOC_TYPE_HEAPINFO) {
369 // We currently do not match fragments with unfragmented blocks (maybe we should).
370 if (heapinfo2b->type == MMALLOC_TYPE_UNFRAGMENTED) {
375 xbt_assert(heapinfo2b->type >= 0, "Unknown mmalloc block type: %d", heapinfo2b->type);
377 for (size_t j2 = 0; j2 < (size_t)(BLOCKSIZE >> heapinfo2b->type); j2++) {
378 if (i2 == i1 && j2 == j1)
381 if (state.equals_to_<2>(i2, j2).valid_)
384 const void* addr_block2 = (ADDR2UINT(i2) - 1) * BLOCKSIZE + (char*)state.std_heap_copy.heapbase;
385 const void* addr_frag2 = (const char*)addr_block2 + (j2 << heapinfo2b->type);
387 if (not heap_area_differ(process, state, addr_frag1, addr_frag2, snapshot1, snapshot2, nullptr, nullptr,
397 XBT_DEBUG("Block %zu, fragment_ %zu not found (size_used = %zd, address = %p)\n", i1, j1,
398 heapinfo1->busy_frag.frag_size[j1], addr_frag1);
406 /* All blocks/fragments are equal to another block/fragment_ ? */
407 for (size_t i = 1; i < state.heaplimit; i++) {
408 const auto* heapinfo1 =
409 static_cast<malloc_info*>(heap_region1->read(&heapinfo_temp1, &heapinfos1[i], sizeof(malloc_info)));
411 if (heapinfo1->type == MMALLOC_TYPE_UNFRAGMENTED && i1 == state.heaplimit && heapinfo1->busy_block.busy_size > 0 &&
412 not state.equals_to_<1>(i, 0).valid_) {
413 XBT_DEBUG("Block %zu not found (size used = %zu)", i, heapinfo1->busy_block.busy_size);
417 if (heapinfo1->type <= 0)
419 for (size_t j = 0; j < (size_t)(BLOCKSIZE >> heapinfo1->type); j++)
420 if (i1 == state.heaplimit && heapinfo1->busy_frag.frag_size[j] > 0 && not state.equals_to_<1>(i, j).valid_) {
421 XBT_DEBUG("Block %zu, Fragment %zu not found (size used = %zd)", i, j, heapinfo1->busy_frag.frag_size[j]);
426 for (size_t i = 1; i < state.heaplimit; i++) {
427 const auto* heapinfo2 =
428 static_cast<malloc_info*>(heap_region2->read(&heapinfo_temp2, &heapinfos2[i], sizeof(malloc_info)));
429 if (heapinfo2->type == MMALLOC_TYPE_UNFRAGMENTED && i1 == state.heaplimit && heapinfo2->busy_block.busy_size > 0 &&
430 not state.equals_to_<2>(i, 0).valid_) {
431 XBT_DEBUG("Block %zu not found (size used = %zu)", i,
432 heapinfo2->busy_block.busy_size);
436 if (heapinfo2->type <= 0)
439 for (size_t j = 0; j < (size_t)(BLOCKSIZE >> heapinfo2->type); j++)
440 if (i1 == state.heaplimit && heapinfo2->busy_frag.frag_size[j] > 0 && not state.equals_to_<2>(i, j).valid_) {
441 XBT_DEBUG("Block %zu, Fragment %zu not found (size used = %zd)",
442 i, j, heapinfo2->busy_frag.frag_size[j]);
452 * @param real_area1 Process address for state 1
453 * @param real_area2 Process address for state 2
454 * @param snapshot1 Snapshot of state 1
455 * @param snapshot2 Snapshot of state 2
458 * @param check_ignore
459 * @return true when different, false otherwise (same or unknown)
461 static bool heap_area_differ_without_type(const RemoteProcessMemory& process, StateComparator& state,
462 const void* real_area1, const void* real_area2, const Snapshot& snapshot1,
463 const Snapshot& snapshot2, HeapLocationPairs* previous, int size,
466 const Region* heap_region1 = MC_get_heap_region(snapshot1);
467 const Region* heap_region2 = MC_get_heap_region(snapshot2);
469 for (int i = 0; i < size; ) {
470 if (check_ignore > 0) {
471 ssize_t ignore1 = heap_comparison_ignore_size(state.processStates[0].to_ignore, (const char*)real_area1 + i);
473 ssize_t ignore2 = heap_comparison_ignore_size(state.processStates[1].to_ignore, (const char*)real_area2 + i);
474 if (ignore2 == ignore1) {
486 if (MC_snapshot_region_memcmp((const char*)real_area1 + i, heap_region1, (const char*)real_area2 + i, heap_region2,
488 int pointer_align = (i / sizeof(void *)) * sizeof(void *);
489 const void* addr_pointed1 = snapshot1.read(remote((void* const*)((const char*)real_area1 + pointer_align)));
490 const void* addr_pointed2 = snapshot2.read(remote((void* const*)((const char*)real_area2 + pointer_align)));
492 if (process.in_maestro_stack(remote(addr_pointed1)) && process.in_maestro_stack(remote(addr_pointed2))) {
493 i = pointer_align + sizeof(void *);
497 if (snapshot1.on_heap(addr_pointed1) && snapshot2.on_heap(addr_pointed2)) {
498 // Both addresses are in the heap:
499 if (heap_area_differ(process, state, addr_pointed1, addr_pointed2, snapshot1, snapshot2, previous, nullptr, 0))
501 i = pointer_align + sizeof(void *);
514 * @param real_area1 Process address for state 1
515 * @param real_area2 Process address for state 2
516 * @param snapshot1 Snapshot of state 1
517 * @param snapshot2 Snapshot of state 2
520 * @param area_size either a byte_size or an elements_count (?)
521 * @param check_ignore
522 * @param pointer_level
523 * @return true when different, false otherwise (same or unknown)
525 static bool heap_area_differ_with_type(const simgrid::mc::RemoteProcessMemory& process, StateComparator& state,
526 const void* real_area1, const void* real_area2, const Snapshot& snapshot1,
527 const Snapshot& snapshot2, HeapLocationPairs* previous, const Type* type,
528 int area_size, int check_ignore, int pointer_level)
530 // HACK: This should not happen but in practice, there are some
531 // DW_TAG_typedef without an associated DW_AT_type:
532 //<1><538832>: Abbrev Number: 111 (DW_TAG_typedef)
533 // <538833> DW_AT_name : (indirect string, offset: 0x2292f3): gregset_t
534 // <538837> DW_AT_decl_file : 98
535 // <538838> DW_AT_decl_line : 37
539 if (is_stack(process, real_area1) && is_stack(process, real_area2))
542 if (check_ignore > 0) {
543 ssize_t ignore1 = heap_comparison_ignore_size(state.processStates[0].to_ignore, real_area1);
544 if (ignore1 > 0 && heap_comparison_ignore_size(state.processStates[1].to_ignore, real_area2) == ignore1)
549 const Type* subsubtype;
551 const void* addr_pointed1;
552 const void* addr_pointed2;
554 const Region* heap_region1 = MC_get_heap_region(snapshot1);
555 const Region* heap_region2 = MC_get_heap_region(snapshot2);
557 switch (type->type) {
558 case DW_TAG_unspecified_type:
561 case DW_TAG_base_type:
562 if (not type->name.empty() && type->name == "char") { /* String, hence random (arbitrary ?) size */
563 if (real_area1 == real_area2)
566 return MC_snapshot_region_memcmp(real_area1, heap_region1, real_area2, heap_region2, area_size) != 0;
568 if (area_size != -1 && type->byte_size != area_size)
571 return MC_snapshot_region_memcmp(real_area1, heap_region1, real_area2, heap_region2, type->byte_size) != 0;
574 case DW_TAG_enumeration_type:
575 if (area_size != -1 && type->byte_size != area_size)
577 return MC_snapshot_region_memcmp(real_area1, heap_region1, real_area2, heap_region2, type->byte_size) != 0;
580 case DW_TAG_const_type:
581 case DW_TAG_volatile_type:
582 return heap_area_differ_with_type(process, state, real_area1, real_area2, snapshot1, snapshot2, previous,
583 type->subtype, area_size, check_ignore, pointer_level);
585 case DW_TAG_array_type:
586 subtype = type->subtype;
587 switch (subtype->type) {
588 case DW_TAG_unspecified_type:
591 case DW_TAG_base_type:
592 case DW_TAG_enumeration_type:
593 case DW_TAG_pointer_type:
594 case DW_TAG_reference_type:
595 case DW_TAG_rvalue_reference_type:
596 case DW_TAG_structure_type:
597 case DW_TAG_class_type:
598 case DW_TAG_union_type:
599 if (subtype->full_type)
600 subtype = subtype->full_type;
601 elm_size = subtype->byte_size;
603 // TODO, just remove the type indirection?
604 case DW_TAG_const_type:
606 case DW_TAG_volatile_type:
607 subsubtype = subtype->subtype;
608 if (subsubtype->full_type)
609 subsubtype = subsubtype->full_type;
610 elm_size = subsubtype->byte_size;
615 for (int i = 0; i < type->element_count; i++) {
616 // TODO, add support for variable stride (DW_AT_byte_stride)
617 if (heap_area_differ_with_type(process, state, (const char*)real_area1 + (i * elm_size),
618 (const char*)real_area2 + (i * elm_size), snapshot1, snapshot2, previous,
619 type->subtype, subtype->byte_size, check_ignore, pointer_level))
624 case DW_TAG_reference_type:
625 case DW_TAG_rvalue_reference_type:
626 case DW_TAG_pointer_type:
627 if (type->subtype && type->subtype->type == DW_TAG_subroutine_type) {
628 addr_pointed1 = snapshot1.read(remote((void* const*)real_area1));
629 addr_pointed2 = snapshot2.read(remote((void* const*)real_area2));
630 return (addr_pointed1 != addr_pointed2);
633 if (pointer_level <= 1) {
634 addr_pointed1 = snapshot1.read(remote((void* const*)real_area1));
635 addr_pointed2 = snapshot2.read(remote((void* const*)real_area2));
636 if (snapshot1.on_heap(addr_pointed1) && snapshot2.on_heap(addr_pointed2))
637 return heap_area_differ(process, state, addr_pointed1, addr_pointed2, snapshot1, snapshot2, previous,
638 type->subtype, pointer_level);
640 return (addr_pointed1 != addr_pointed2);
642 for (size_t i = 0; i < (area_size / sizeof(void*)); i++) {
643 addr_pointed1 = snapshot1.read(remote((void* const*)((const char*)real_area1 + i * sizeof(void*))));
644 addr_pointed2 = snapshot2.read(remote((void* const*)((const char*)real_area2 + i * sizeof(void*))));
645 bool differ = snapshot1.on_heap(addr_pointed1) && snapshot2.on_heap(addr_pointed2)
646 ? heap_area_differ(process, state, addr_pointed1, addr_pointed2, snapshot1, snapshot2,
647 previous, type->subtype, pointer_level)
648 : addr_pointed1 != addr_pointed2;
654 case DW_TAG_structure_type:
655 case DW_TAG_class_type:
657 type = type->full_type;
658 if (type->byte_size == 0)
660 if (area_size != -1 && type->byte_size != area_size) {
661 if (area_size <= type->byte_size || area_size % type->byte_size != 0)
663 for (size_t i = 0; i < (size_t)(area_size / type->byte_size); i++) {
664 if (heap_area_differ_with_type(process, state, (const char*)real_area1 + i * type->byte_size,
665 (const char*)real_area2 + i * type->byte_size, snapshot1, snapshot2, previous,
666 type, -1, check_ignore, 0))
670 for (const simgrid::mc::Member& member : type->members) {
671 // TODO, optimize this? (for the offset case)
672 const void* real_member1 = dwarf::resolve_member(real_area1, type, &member, &snapshot1);
673 const void* real_member2 = dwarf::resolve_member(real_area2, type, &member, &snapshot2);
674 if (heap_area_differ_with_type(process, state, real_member1, real_member2, snapshot1, snapshot2, previous,
675 member.type, -1, check_ignore, 0))
681 case DW_TAG_union_type:
682 return heap_area_differ_without_type(process, state, real_area1, real_area2, snapshot1, snapshot2, previous,
683 type->byte_size, check_ignore);
690 /** Infer the type of a part of the block from the type of the block
692 * TODO, handle DW_TAG_array_type as well as arrays of the object ((*p)[5], p[5])
694 * TODO, handle subfields ((*p).bar.foo, (*p)[5].bar…)
696 * @param type DWARF type ID of the root address
698 * @return DWARF type ID for given offset
700 static Type* get_offset_type(void* real_base_address, Type* type, int offset, int area_size, const Snapshot& snapshot)
702 // Beginning of the block, the inferred variable type if the type of the block:
706 switch (type->type) {
707 case DW_TAG_structure_type:
708 case DW_TAG_class_type:
710 type = type->full_type;
711 if (area_size != -1 && type->byte_size != area_size) {
712 if (area_size > type->byte_size && area_size % type->byte_size == 0)
718 for (const simgrid::mc::Member& member : type->members) {
719 if (member.has_offset_location()) {
720 // We have the offset, use it directly (shortcut):
721 if (member.offset() == offset)
724 void* real_member = dwarf::resolve_member(real_base_address, type, &member, &snapshot);
725 if ((char*)real_member - (char*)real_base_address == offset)
732 /* FIXME: other cases ? */
739 * @param area1 Process address for state 1
740 * @param area2 Process address for state 2
741 * @param snapshot1 Snapshot of state 1
742 * @param snapshot2 Snapshot of state 2
743 * @param previous Pairs of blocks already compared on the current path (or nullptr)
744 * @param type_id Type of variable
745 * @param pointer_level
746 * @return true when different, false otherwise (same or unknown)
748 static bool heap_area_differ(const RemoteProcessMemory& process, StateComparator& state, const void* area1,
749 const void* area2, const Snapshot& snapshot1, const Snapshot& snapshot2,
750 HeapLocationPairs* previous, Type* type, int pointer_level)
755 int check_ignore = 0;
763 Type* new_type1 = nullptr;
765 bool match_pairs = false;
767 // This is the address of std_heap->heapinfo in the application process:
768 uint64_t heapinfo_address = process.heap_address.address() + offsetof(s_xbt_mheap_t, heapinfo);
770 const malloc_info* heapinfos1 = snapshot1.read(remote<malloc_info*>(heapinfo_address));
771 const malloc_info* heapinfos2 = snapshot2.read(remote<malloc_info*>(heapinfo_address));
773 malloc_info heapinfo_temp1;
774 malloc_info heapinfo_temp2;
776 simgrid::mc::HeapLocationPairs current;
777 if (previous == nullptr) {
783 block1 = ((const char*)area1 - (const char*)state.std_heap_copy.heapbase) / BLOCKSIZE + 1;
784 block2 = ((const char*)area2 - (const char*)state.std_heap_copy.heapbase) / BLOCKSIZE + 1;
786 // If either block is a stack block:
787 if (is_block_stack(process, (int)block1) && is_block_stack(process, (int)block2)) {
788 previous->insert(HeapLocationPair{{HeapLocation(block1, -1), HeapLocation(block2, -1)}});
790 state.match_equals(previous);
794 // If either block is not in the expected area of memory:
795 if (((const char*)area1 < (const char*)state.std_heap_copy.heapbase) ||
796 (block1 > (ssize_t)state.processStates[0].heapsize) ||
797 ((const char*)area2 < (const char*)state.std_heap_copy.heapbase) ||
798 (block2 > (ssize_t)state.processStates[1].heapsize)) {
802 // Process address of the block:
803 void* real_addr_block1 = (ADDR2UINT(block1) - 1) * BLOCKSIZE + (char*)state.std_heap_copy.heapbase;
804 void* real_addr_block2 = (ADDR2UINT(block2) - 1) * BLOCKSIZE + (char*)state.std_heap_copy.heapbase;
808 type = type->full_type;
810 // This assume that for "boring" types (volatile ...) byte_size is absent:
811 while (type->byte_size == 0 && type->subtype != nullptr)
812 type = type->subtype;
815 if (type->type == DW_TAG_pointer_type ||
816 (type->type == DW_TAG_base_type && not type->name.empty() && type->name == "char"))
819 type_size = type->byte_size;
822 const Region* heap_region1 = MC_get_heap_region(snapshot1);
823 const Region* heap_region2 = MC_get_heap_region(snapshot2);
825 const auto* heapinfo1 =
826 static_cast<malloc_info*>(heap_region1->read(&heapinfo_temp1, &heapinfos1[block1], sizeof(malloc_info)));
827 const auto* heapinfo2 =
828 static_cast<malloc_info*>(heap_region2->read(&heapinfo_temp2, &heapinfos2[block2], sizeof(malloc_info)));
830 if ((heapinfo1->type == MMALLOC_TYPE_FREE || heapinfo1->type==MMALLOC_TYPE_HEAPINFO)
831 && (heapinfo2->type == MMALLOC_TYPE_FREE || heapinfo2->type ==MMALLOC_TYPE_HEAPINFO)) {
834 state.match_equals(previous);
838 if (heapinfo1->type == MMALLOC_TYPE_UNFRAGMENTED && heapinfo2->type == MMALLOC_TYPE_UNFRAGMENTED) {
841 // TODO, lookup variable type from block type as done for fragmented blocks
843 if (state.equals_to_<1>(block1, 0).valid_ && state.equals_to_<2>(block2, 0).valid_ &&
844 state.blocksEqual(block1, block2)) {
846 state.match_equals(previous);
850 if (type_size != -1 && type_size != (ssize_t)heapinfo1->busy_block.busy_size &&
851 type_size != (ssize_t)heapinfo2->busy_block.busy_size && type->name.empty()) {
853 state.match_equals(previous);
857 if (heapinfo1->busy_block.size != heapinfo2->busy_block.size ||
858 heapinfo1->busy_block.busy_size != heapinfo2->busy_block.busy_size)
861 if (not previous->insert(HeapLocationPair{{HeapLocation(block1, -1), HeapLocation(block2, -1)}}).second) {
863 state.match_equals(previous);
867 size = heapinfo1->busy_block.busy_size;
869 // Remember (basic) type inference.
870 // The current data structure only allows us to do this for the whole block.
871 if (type != nullptr && area1 == real_addr_block1)
872 state.types_<1>(block1, 0) = type;
873 if (type != nullptr && area2 == real_addr_block2)
874 state.types_<2>(block2, 0) = type;
878 state.match_equals(previous);
882 if (heapinfo1->busy_block.ignore > 0 && heapinfo2->busy_block.ignore == heapinfo1->busy_block.ignore)
883 check_ignore = heapinfo1->busy_block.ignore;
885 } else if ((heapinfo1->type > 0) && (heapinfo2->type > 0)) { /* Fragmented block */
887 ssize_t frag1 = (ADDR2UINT(area1) % BLOCKSIZE) >> heapinfo1->type;
888 ssize_t frag2 = (ADDR2UINT(area2) % BLOCKSIZE) >> heapinfo2->type;
890 // Process address of the fragment_:
891 void* real_addr_frag1 = (char*)real_addr_block1 + (frag1 << heapinfo1->type);
892 void* real_addr_frag2 = (char*)real_addr_block2 + (frag2 << heapinfo2->type);
894 // Check the size of the fragments against the size of the type:
895 if (type_size != -1) {
896 if (heapinfo1->busy_frag.frag_size[frag1] == -1 || heapinfo2->busy_frag.frag_size[frag2] == -1) {
898 state.match_equals(previous);
902 if (type_size != heapinfo1->busy_frag.frag_size[frag1]
903 || type_size != heapinfo2->busy_frag.frag_size[frag2]) {
905 state.match_equals(previous);
910 // Check if the blocks are already matched together:
911 if (state.equals_to_<1>(block1, frag1).valid_ && state.equals_to_<2>(block2, frag2).valid_ &&
912 state.fragmentsEqual(block1, frag1, block2, frag2)) {
914 state.match_equals(previous);
917 // Compare the size of both fragments:
918 if (heapinfo1->busy_frag.frag_size[frag1] != heapinfo2->busy_frag.frag_size[frag2]) {
919 if (type_size == -1) {
921 state.match_equals(previous);
927 // Size of the fragment_:
928 size = heapinfo1->busy_frag.frag_size[frag1];
930 // Remember (basic) type inference.
931 // The current data structure only allows us to do this for the whole fragment_.
932 if (type != nullptr && area1 == real_addr_frag1)
933 state.types_<1>(block1, frag1) = type;
934 if (type != nullptr && area2 == real_addr_frag2)
935 state.types_<2>(block2, frag2) = type;
937 // The type of the variable is already known:
941 // Type inference from the block type.
942 else if (state.types_<1>(block1, frag1) != nullptr || state.types_<2>(block2, frag2) != nullptr) {
943 Type* new_type2 = nullptr;
945 offset1 = (const char*)area1 - (const char*)real_addr_frag1;
946 offset2 = (const char*)area2 - (const char*)real_addr_frag2;
948 if (state.types_<1>(block1, frag1) != nullptr && state.types_<2>(block2, frag2) != nullptr) {
949 new_type1 = get_offset_type(real_addr_frag1, state.types_<1>(block1, frag1), offset1, size, snapshot1);
950 new_type2 = get_offset_type(real_addr_frag2, state.types_<2>(block2, frag2), offset1, size, snapshot2);
951 } else if (state.types_<1>(block1, frag1) != nullptr) {
952 new_type1 = get_offset_type(real_addr_frag1, state.types_<1>(block1, frag1), offset1, size, snapshot1);
953 new_type2 = get_offset_type(real_addr_frag2, state.types_<1>(block1, frag1), offset2, size, snapshot2);
954 } else if (state.types_<2>(block2, frag2) != nullptr) {
955 new_type1 = get_offset_type(real_addr_frag1, state.types_<2>(block2, frag2), offset1, size, snapshot1);
956 new_type2 = get_offset_type(real_addr_frag2, state.types_<2>(block2, frag2), offset2, size, snapshot2);
959 state.match_equals(previous);
963 if (new_type1 != nullptr && new_type2 != nullptr && new_type1 != new_type2) {
965 while (type->byte_size == 0 && type->subtype != nullptr)
966 type = type->subtype;
967 new_size1 = type->byte_size;
970 while (type->byte_size == 0 && type->subtype != nullptr)
971 type = type->subtype;
972 new_size2 = type->byte_size;
976 state.match_equals(previous);
981 if (new_size1 > 0 && new_size1 == new_size2) {
986 if (offset1 == 0 && offset2 == 0 &&
987 not previous->insert(HeapLocationPair{{HeapLocation(block1, frag1), HeapLocation(block2, frag2)}}).second) {
989 state.match_equals(previous);
995 state.match_equals(previous);
999 if ((heapinfo1->busy_frag.ignore[frag1] > 0) &&
1000 (heapinfo2->busy_frag.ignore[frag2] == heapinfo1->busy_frag.ignore[frag1]))
1001 check_ignore = heapinfo1->busy_frag.ignore[frag1];
1005 /* Start comparison */
1006 if (type ? heap_area_differ_with_type(process, state, area1, area2, snapshot1, snapshot2, previous, type, size,
1007 check_ignore, pointer_level)
1008 : heap_area_differ_without_type(process, state, area1, area2, snapshot1, snapshot2, previous, size,
1013 state.match_equals(previous);
1016 } // namespace simgrid::mc
1018 /************************** Snapshot comparison *******************************/
1019 /******************************************************************************/
1021 static bool areas_differ_with_type(const simgrid::mc::RemoteProcessMemory& process, simgrid::mc::StateComparator& state,
1022 const void* real_area1, const simgrid::mc::Snapshot& snapshot1,
1023 simgrid::mc::Region* region1, const void* real_area2,
1024 const simgrid::mc::Snapshot& snapshot2, simgrid::mc::Region* region2,
1025 const simgrid::mc::Type* type, int pointer_level)
1027 const simgrid::mc::Type* subtype;
1028 const simgrid::mc::Type* subsubtype;
1031 xbt_assert(type != nullptr);
1032 switch (type->type) {
1033 case DW_TAG_unspecified_type:
1036 case DW_TAG_base_type:
1037 case DW_TAG_enumeration_type:
1038 case DW_TAG_union_type:
1039 return MC_snapshot_region_memcmp(real_area1, region1, real_area2, region2, type->byte_size) != 0;
1040 case DW_TAG_typedef:
1041 case DW_TAG_volatile_type:
1042 case DW_TAG_const_type:
1043 return areas_differ_with_type(process, state, real_area1, snapshot1, region1, real_area2, snapshot2, region2,
1044 type->subtype, pointer_level);
1045 case DW_TAG_array_type:
1046 subtype = type->subtype;
1047 switch (subtype->type) {
1048 case DW_TAG_unspecified_type:
1051 case DW_TAG_base_type:
1052 case DW_TAG_enumeration_type:
1053 case DW_TAG_pointer_type:
1054 case DW_TAG_reference_type:
1055 case DW_TAG_rvalue_reference_type:
1056 case DW_TAG_structure_type:
1057 case DW_TAG_class_type:
1058 case DW_TAG_union_type:
1059 if (subtype->full_type)
1060 subtype = subtype->full_type;
1061 elm_size = subtype->byte_size;
1063 case DW_TAG_const_type:
1064 case DW_TAG_typedef:
1065 case DW_TAG_volatile_type:
1066 subsubtype = subtype->subtype;
1067 if (subsubtype->full_type)
1068 subsubtype = subsubtype->full_type;
1069 elm_size = subsubtype->byte_size;
1074 for (int i = 0; i < type->element_count; i++) {
1075 size_t off = i * elm_size;
1076 if (areas_differ_with_type(process, state, (const char*)real_area1 + off, snapshot1, region1,
1077 (const char*)real_area2 + off, snapshot2, region2, type->subtype, pointer_level))
1081 case DW_TAG_pointer_type:
1082 case DW_TAG_reference_type:
1083 case DW_TAG_rvalue_reference_type: {
1084 const void* addr_pointed1 = MC_region_read_pointer(region1, real_area1);
1085 const void* addr_pointed2 = MC_region_read_pointer(region2, real_area2);
1087 if (type->subtype && type->subtype->type == DW_TAG_subroutine_type)
1088 return (addr_pointed1 != addr_pointed2);
1089 if (addr_pointed1 == nullptr && addr_pointed2 == nullptr)
1091 if (addr_pointed1 == nullptr || addr_pointed2 == nullptr)
1093 if (not state.compared_pointers.insert(std::make_pair(addr_pointed1, addr_pointed2)).second)
1098 // Some cases are not handled here:
1099 // * the pointers lead to different areas (one to the heap, the other to the RW segment ...)
1100 // * a pointer leads to the read-only segment of the current object
1101 // * a pointer lead to a different ELF object
1103 if (snapshot1.on_heap(addr_pointed1)) {
1104 if (not snapshot2.on_heap(addr_pointed2))
1106 // The pointers are both in the heap:
1107 return simgrid::mc::heap_area_differ(process, state, addr_pointed1, addr_pointed2, snapshot1, snapshot2,
1108 nullptr, type->subtype, pointer_level);
1110 } else if (region1->contain(simgrid::mc::remote(addr_pointed1))) {
1111 // The pointers are both in the current object R/W segment:
1112 if (not region2->contain(simgrid::mc::remote(addr_pointed2)))
1114 if (not type->type_id)
1115 return (addr_pointed1 != addr_pointed2);
1117 return areas_differ_with_type(process, state, addr_pointed1, snapshot1, region1, addr_pointed2, snapshot2,
1118 region2, type->subtype, pointer_level);
1120 // TODO, We do not handle very well the case where
1121 // it belongs to a different (non-heap) region from the current one.
1123 return (addr_pointed1 != addr_pointed2);
1126 case DW_TAG_structure_type:
1127 case DW_TAG_class_type:
1128 for (const simgrid::mc::Member& member : type->members) {
1129 const void* member1 = simgrid::dwarf::resolve_member(real_area1, type, &member, &snapshot1);
1130 const void* member2 = simgrid::dwarf::resolve_member(real_area2, type, &member, &snapshot2);
1131 simgrid::mc::Region* subregion1 = snapshot1.get_region(member1, region1); // region1 is hinted
1132 simgrid::mc::Region* subregion2 = snapshot2.get_region(member2, region2); // region2 is hinted
1133 if (areas_differ_with_type(process, state, member1, snapshot1, subregion1, member2, snapshot2, subregion2,
1134 member.type, pointer_level))
1138 case DW_TAG_subroutine_type:
1141 XBT_VERB("Unknown case: %d", type->type);
1148 static bool global_variables_differ(const simgrid::mc::RemoteProcessMemory& process,
1149 simgrid::mc::StateComparator& state,
1150 const simgrid::mc::ObjectInformation* object_info, simgrid::mc::Region* r1,
1151 simgrid::mc::Region* r2, const simgrid::mc::Snapshot& snapshot1,
1152 const simgrid::mc::Snapshot& snapshot2)
1154 xbt_assert(r1 && r2, "Missing region.");
1156 const std::vector<simgrid::mc::Variable>& variables = object_info->global_variables;
1158 for (simgrid::mc::Variable const& current_var : variables) {
1159 // If the variable is not in this object, skip it:
1160 // We do not expect to find a pointer to something which is not reachable
1161 // by the global variables.
1162 if ((char*)current_var.address < object_info->start_rw || (char*)current_var.address > object_info->end_rw)
1165 const simgrid::mc::Type* bvariable_type = current_var.type;
1166 if (areas_differ_with_type(process, state, current_var.address, snapshot1, r1, current_var.address, snapshot2, r2,
1167 bvariable_type, 0)) {
1168 XBT_VERB("Global variable %s (%p) is different between snapshots", current_var.name.c_str(), current_var.address);
1176 static bool local_variables_differ(const simgrid::mc::RemoteProcessMemory& process, simgrid::mc::StateComparator& state,
1177 const simgrid::mc::Snapshot& snapshot1, const simgrid::mc::Snapshot& snapshot2,
1178 const_mc_snapshot_stack_t stack1, const_mc_snapshot_stack_t stack2)
1180 if (stack1->local_variables.size() != stack2->local_variables.size()) {
1181 XBT_VERB("Different number of local variables");
1185 for (unsigned int cursor = 0; cursor < stack1->local_variables.size(); cursor++) {
1186 const_local_variable_t current_var1 = &stack1->local_variables[cursor];
1187 const_local_variable_t current_var2 = &stack2->local_variables[cursor];
1188 if (current_var1->name != current_var2->name || current_var1->subprogram != current_var2->subprogram ||
1189 current_var1->ip != current_var2->ip) {
1190 // TODO, fix current_varX->subprogram->name to include name if DW_TAG_inlined_subprogram
1191 XBT_VERB("Different name of variable (%s - %s) or frame (%s - %s) or ip (%lu - %lu)", current_var1->name.c_str(),
1192 current_var2->name.c_str(), current_var1->subprogram->name.c_str(),
1193 current_var2->subprogram->name.c_str(), current_var1->ip, current_var2->ip);
1197 if (areas_differ_with_type(process, state, current_var1->address, snapshot1,
1198 snapshot1.get_region(current_var1->address), current_var2->address, snapshot2,
1199 snapshot2.get_region(current_var2->address), current_var1->type, 0)) {
1200 XBT_VERB("Local variable %s (%p - %p) in frame %s is different between snapshots", current_var1->name.c_str(),
1201 current_var1->address, current_var2->address, current_var1->subprogram->name.c_str());
1208 namespace simgrid::mc {
1210 bool Snapshot::operator==(const Snapshot& other)
1212 // TODO, make this a field of ModelChecker or something similar
1213 static StateComparator state_comparator;
1215 RemoteProcessMemory& memory = mc_model_checker->get_remote_process_memory();
1217 if (hash_ != other.hash_) {
1218 XBT_VERB("(%ld - %ld) Different hash: 0x%" PRIx64 "--0x%" PRIx64, this->num_state_, other.num_state_, this->hash_,
1222 XBT_VERB("(%ld - %ld) Same hash: 0x%" PRIx64, this->num_state_, other.num_state_, this->hash_);
1224 /* TODO: re-enable the quick filter of counting enabled processes in each snapshots */
1226 /* Compare size of stacks */
1227 for (unsigned long i = 0; i < this->stacks_.size(); i++) {
1228 size_t size_used1 = this->stack_sizes_[i];
1229 size_t size_used2 = other.stack_sizes_[i];
1230 if (size_used1 != size_used2) {
1231 XBT_VERB("(%ld - %ld) Different size used in stacks: %zu - %zu", num_state_, other.num_state_, size_used1,
1237 /* Init heap information used in heap comparison algorithm */
1238 const s_xbt_mheap_t* heap1 = static_cast<xbt_mheap_t>(
1239 this->read_bytes(alloca(sizeof(s_xbt_mheap_t)), sizeof(s_xbt_mheap_t), memory.heap_address, ReadOptions::lazy()));
1240 const s_xbt_mheap_t* heap2 = static_cast<xbt_mheap_t>(
1241 other.read_bytes(alloca(sizeof(s_xbt_mheap_t)), sizeof(s_xbt_mheap_t), memory.heap_address, ReadOptions::lazy()));
1242 if (state_comparator.initHeapInformation(memory, heap1, heap2, this->to_ignore_, other.to_ignore_) == -1) {
1243 XBT_VERB("(%ld - %ld) Different heap information", this->num_state_, other.num_state_);
1247 /* Stacks comparison */
1248 for (unsigned int cursor = 0; cursor < this->stacks_.size(); cursor++) {
1249 const_mc_snapshot_stack_t stack1 = &this->stacks_[cursor];
1250 const_mc_snapshot_stack_t stack2 = &other.stacks_[cursor];
1252 if (local_variables_differ(memory, state_comparator, *this, other, stack1, stack2)) {
1253 XBT_VERB("(%ld - %ld) Different local variables between stacks %u", this->num_state_, other.num_state_,
1259 size_t regions_count = this->snapshot_regions_.size();
1260 if (regions_count != other.snapshot_regions_.size())
1263 for (size_t k = 0; k != regions_count; ++k) {
1264 Region* region1 = this->snapshot_regions_[k].get();
1265 Region* region2 = other.snapshot_regions_[k].get();
1268 if (region1->region_type() != RegionType::Data)
1271 xbt_assert(region1->region_type() == region2->region_type());
1272 xbt_assert(region1->object_info() == region2->object_info());
1273 xbt_assert(region1->object_info());
1275 /* Compare global variables */
1276 if (global_variables_differ(memory, state_comparator, region1->object_info(), region1, region2, *this, other)) {
1277 std::string const& name = region1->object_info()->file_name;
1278 XBT_VERB("(%ld - %ld) Different global variables in %s", this->num_state_, other.num_state_, name.c_str());
1283 XBT_VERB(" Compare heap...");
1285 if (mmalloc_heap_differ(memory, state_comparator, *this, other)) {
1286 XBT_VERB("(%ld - %ld) Different heap (mmalloc_heap_differ)", this->num_state_, other.num_state_);
1290 XBT_VERB("(%ld - %ld) No difference found", this->num_state_, other.num_state_);
1294 } // namespace simgrid::mc