Fix for race condition in rosbag2 recording tests with wait_for_db()

[MichaelOrlov]

• Closes Race condition in rosbag2 recording tests with wait_for_db() #1140
• Getting rid from attempt to open DB file in wait_for_db() test infrastructure to avoid race condition with rosbag2 recorder.

RCA

After some analysis @MichaelOrlov came to the conclusion that failure happening due to the race condition for the resource. i.e. database file.
In recording integration tests we usually spawn child process with ros2 bag record command and in parent process calling wait_for_db() function.
The problem is that wait_for_db() inside trying to open the same database file as recorder supposed to create for recording.

rosbag2/rosbag2_tests/test/rosbag2_tests/record_fixture.hpp

Lines 120 to 136 in 3df2fe3

	void wait_for_db()
	{
	const auto database_path = get_bag_file_path(0);
	while (true) {
	try {
	std::this_thread::sleep_for(50ms); // wait a bit to not query constantly
	if (database_path.exists()) {
	rosbag2_storage_plugins::SqliteWrapper db{
	database_path.string(), rosbag2_storage::storage_interfaces::IOFlag::READ_ONLY};
	return;
	}
	} catch (const rosbag2_storage_plugins::SqliteException & ex) {
	(void) ex; // still waiting
	}
	}
	}

Some times wait_for_db() creating DB reader and locking database at the moment when recorder already created database but hasn't fully initialized it.
Somewhere in between one of the database_->prepare_statement(create_stmt)->execute_and_reset();

rosbag2/rosbag2_storage_sqlite3/src/rosbag2_storage_sqlite3/sqlite_storage.cpp

Lines 377 to 385 in 3df2fe3

	database_->prepare_statement(create_stmt)->execute_and_reset();
	create_stmt = "CREATE TABLE messages(" \
	"id INTEGER PRIMARY KEY," \
	"topic_id INTEGER NOT NULL," \
	"timestamp INTEGER NOT NULL, " \
	"data BLOB NOT NULL);";
	database_->prepare_statement(create_stmt)->execute_and_reset();
	create_stmt = "CREATE INDEX timestamp_idx ON messages (timestamp ASC);";
	database_->prepare_statement(create_stmt)->execute_and_reset();

Recorder fails in this case since database get occupied from another parent process.

Signed-off-by: Michael Orlov michael.orlov@apex.ai

[MichaelOrlov]

Gist: https://gist.githubusercontent.com/MichaelOrlov/7b93540c8d427e4e59d775efd271a784/raw/542cc78201056376b2637c961bca22da156ba526/ros2.repos
BUILD args: --packages-above-and-dependencies rosbag2_tests rosbag2_transport
TEST args: --packages-above rosbag2_tests rosbag2_transport
ROS Distro: rolling
Job: ci_launcher
ci_launcher ran: https://ci.ros2.org/job/ci_launcher/11059

• Linux
• Linux-aarch64
• Windows

[MichaelOrlov]

https://github.com/Mergifyio backport humble

[mergify]

backport humble

✅ Backports have been created

• #1148 Fix for race condition in rosbag2 recording tests with wait_for_db() (backport #1141) has been created for branch humble