MySQL defines a macro log_msg in one of its headers, which
causes log_msg to be replaced with different symbol in
some compilation units.
Fix by renaming the struct to log_msg_t to avoid the conflict.
The thread local ti_thread object was not initialized for main thread,
which caused ASAN to complain about wild pointer.
Fixed by assigning a thread local ti_thread object for main
thread too.
Avoid streaming context cleanup in after_rollback() hook, if
transaction is going to replay. Otherwise, streaming client is not
stopped and is left in streaming client map, potentially causing
"Failed to insert streaming client" warning if the transaction
goes for retry.
Remove bf abort handling `client_state::after_statement()`, since the
same logic already appears later in `transaction::after_statement()`.
Also, introduce `transaction::after_statement()` overload which takes
a lock.
In case if SR transaction gets BF aborted before it is
certified, the replay will happen but if the replay fails
due to certification conflict, the transaction is not terminated
in provider replay and will keep its streaming status.
Removed the wrong assertion about streaming status from
transaction::cleanup().
Allocate separate client state in mock_client_service::replay()
for replaying step.
Added two new test cases for streaming replication replay after
BF abort.
* Move resetting is_bf_immutable_ into trasaction::cleanup()
to ensure that it is reset to false regardless how the
transaction terminates.
* Removed redundant lock()/unlock() methods from
mock_client_state.
The transaction state is set to s_ordered_commit in
ordered_commit(). However, this is too late for making the
transaction immune for BF aborts after commit order has
been established, which happens in before_commit().
Moving the state change into before_commit() would be the
right thing to do, but that would require too many fixes
to existing applications which are using the lib.
In order to make the transaction immune for BF abort
after it has been ordered to commit, introduce additional
boolean flag which is set to true at the end of before_commit()
and is taken into account in bf_abort().
Make is_rollbacker_active() public so that the BF thread can
check if the rollbacker was started or not.
Also don't unlock the lock for launching the background
rollbacker to avoid race conditions in accessing the
victim state.
Replace ubuntu-18.04 virtual environment with ubuntu-20.04.
Compilation tests for GCC < 7 cannot be done anymore
since ubuntu-20.04 environment does not provide such an
old compilers.
Added methods bf_abort() and total_order_bf_abort() which take
wsrep::unique_lock<wsrep::mutex> as argument to allow caller
to grab the mutex before attempting BF abort. The old calls
were kept for backwards compatibility and wrap the new calls
with internal locking.
Streaming_context was cleaned up in streaming_rollback(),
which could cause clearing fragment seqno vector while
it was still accessed by the owning thread, causing
undefined behavior. Fixed by postponing streaming_context
cleanup for BF aborted SR transactions to always happen in
after_rollback().
Streaming rollback for total order BF abort used regular
BF abort codepath, which was not correct because the streaming
rollback must fully complete before total order operation executes.
Fixed this by adjusting bf_aborted_in_total_order_
before streaming_rollback() gets called.
The condition to skip changing to `s_joined` for all codepaths
which return from donor state. Extracted the logic into separate
method.
Commented start_sst_action in mock_server_service.
When the donor lost its donor state during SST due to cluster
partitioning, the state was erranously changed to `s_joined`
in `start_sst()` and `sst_sent()`, which caused assertion failures
in state checking.
Fixed by changing state to `s_joined` only if donor is still in
`s_donor` state.
Use pointers to pass state objects to service constructors
to work around GCC 12 warning
error: member ‘wsrep::mock_storage_service::client_state_’
is used uninitialized
Removed calls to assert() from public headers to have
full control when assertions are enabled in wsrep-lib
code regardless of parent project build configuration.
Moved methods containing assertions and non-trivial
code from headers into compilation units.
Compilation failed on arm64 with
error: comparison is always true due to limited range of data type
[-Werror=type-limits]
for
if (0x0 <= *c && *c <= 0x1f)
This was because char is unsigned on arm64 and thus always greater
than zero.
Fixed by using std::iscntrl() instead of explicit range check.
This adds also backspace into set of escaped characters.
Changed server_state public methods sst_received() and wait_until_state()
to report errors as return value instead of throwing exceptions.
This was done to gradually get rid of public methods which report
errors via exceptions.
This change was part of MDEV-30419.
Report event will write json formatted event into report
file.
Include Boost headers as system headers to avoid generating
excessive warnings. Enable extra tests for selected compilers
in actions.
Grabbing back the lock later, after set_position has been called.
This is because set_position may have to wait for correct seqno position
and calls sync wait in galera side. Such wait would happen while holding
the lock, which would case hanging like reported in MDEV-29512
PR for MDEV-29512 contains a mtr test for reproducing one such deadlock
scenario.
- Change default WITH_BOOST path to reside inside source tree
to avoid permission problems or accidentally overwriting
directories outside source tree.
- If system Boost installation is not found, optionally use
header only unit test framework from location specified by
WITH_BOOST.
- Fix extra-semi warning in provider_options.hpp.
This commit introduces provider_options module which
acts as a proxy between application and wsrep provider
library.
When initialized, the provider options are read from the
provider through config service api extension. A lookup
table with current values and defaults is constructed.
Parameter names are mapped so that dots in provider parameter
names are transformed to underscores.
It turns out that avoiding apply error on fragment removal failure, is
not a safe thing to do. If the DBMS restarts, with a entry in the
streaming log storage, it may be recovered by creating a corresponding
streaming applier.
This reverts commit da5098b622bc6f92c9265bacf4c1168fd69b38b5.
Return an error if server_id_ is undefined before fragment
certification. This may happen if the server disconnects from the
cluster, right before a transactions attempts to replicate a
fragment.
Attempt to append keys while transaction is not active results in
creating a transaction handle with id -1 (undefined).
Assert that the transaction is `active()` before appending keys.
Same for appending data.
