Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

持久化代码合并double write buffer后测试 #8

Merged
merged 4 commits into from
Apr 18, 2024

Conversation

hnwyllmm
Copy link
Owner

What problem were solved in this pull request?

Problem:
合并double write buffer代码后,一些单元测试无法测试通过

主要修复了Bplus tree log测试:当一个frame涉及多个动作时,只重做了第一个动作(这个与dwb无关);
bplus tree 重做没有unpin frame;
double write buffer 导致数据没有完全写入磁盘,而写入到了共享dwb文件中。

@hnwyllmm hnwyllmm merged commit 90bee58 into durability Apr 18, 2024
hnwyllmm pushed a commit that referenced this pull request Oct 22, 2024
### What problem were solved in this pull request?

**Problem:**

When I attempted to join multiple tables, `libsanitizer` detected a
memory leak. The error log is as follows:

```
==23682==ERROR: LeakSanitizer: detected memory leaks

Direct leak of 768 byte(s) in 12 object(s) allocated from:
    #0 0x75f738946a2a in operator new(unsigned long) /usr/src/debug/gcc/gcc/libsanitizer/asan/asan_new_delete.cpp:95
    #1 0x56870207d56f in RowTuple::set_schema(Table const*, std::vector<FieldMeta, std::allocator<FieldMeta> > const*) /home/pepefloyd/git/miniob/src/observer/sql/expr/tuple.h:184
    #2 0x56870208f737 in TableScanPhysicalOperator::open(Trx*) /home/pepefloyd/git/miniob/src/observer/sql/operator/table_scan_physical_operator.cpp:25
    #3 0x568702081888 in NestedLoopJoinPhysicalOperator::right_next() /home/pepefloyd/git/miniob/src/observer/sql/operator/join_physical_operator.cpp:113
    #4 0x568702083b42 in PredicatePhysicalOperator::next() /home/pepefloyd/git/miniob/src/observer/sql/operator/predicate_physical_operator.cpp:41
    #5 0x56870206f411 in HashGroupByPhysicalOperator::open(Trx*) /home/pepefloyd/git/miniob/src/observer/sql/operator/hash_group_by_physical_operator.cpp:44
    #6 0x568702085740 in ProjectPhysicalOperator::open(Trx*) /home/pepefloyd/git/miniob/src/observer/sql/operator/project_physical_operator.cpp:34
    #7 0x568702038e2c in PlainCommunicator::write_result_internal(SessionEvent*, bool&) /home/pepefloyd/git/miniob/src/observer/net/plain_communicator.cpp:193
    #8 0x56870203a7cc in PlainCommunicator::write_result(SessionEvent*, bool&) /home/pepefloyd/git/miniob/src/observer/net/plain_communicator.cpp:162
    #9 0x568701ee5e4b in SqlTaskHandler::handle_event(Communicator*) /home/pepefloyd/git/miniob/src/observer/net/sql_task_handler.cpp:45
    #10 0x56870202f0d7 in Worker::operator()() /home/pepefloyd/git/miniob/src/observer/net/one_thread_per_connection_thread_handler.cpp:96
    #11 0x75f7384f78b3 in execute_native_thread_routine /usr/src/debug/gcc/gcc/libstdc++-v3/src/c++11/thread.cc:104

SUMMARY: AddressSanitizer: 768 byte(s) leaked in 12 allocation(s).
```

### What is changed and how it works?

To fix the memory leak, I identified that in the function
`Tuple::set_schema`:

```cpp
void set_schema(const Table *table, const std::vector<FieldMeta> *fields)
{
  table_ = table;
  // fix: During a join, the right table's open is called multiple times,
  // which invokes set_schema repeatedly. This causes Tuple to store
  // unnecessary fields and values, so it is necessary to clear the existing ones first.
  this->speces_.clear();
  this->speces_.reserve(fields->size());
  for (const FieldMeta &field : *fields) {
    speces_.push_back(new FieldExpr(table, &field));
  }
}
```

The call to `this->speces_.clear()` does not destruct the existing
`FieldExpr*` objects. As a result, the memory was not released properly.
The fix I applied is to manually delete the `FieldExpr*` before clearing
the vector, as shown below:

```cpp
for (FieldExpr *spec : speces_) {
  delete spec;
}
```

This ensures that memory allocated for `FieldExpr` is properly released,
thus fixing the memory leak.
@hnwyllmm hnwyllmm deleted the durability-test branch October 30, 2024 06:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant