6
0
mirror of https://github.com/FirebirdSQL/firebird-qa.git synced 2025-01-22 13:33:07 +01:00
firebird-qa/tests/bugs/core_2988_test.py

261 lines
8.8 KiB
Python

#coding:utf-8
"""
ID: issue-3370
ISSUE: 3370
TITLE: Concurrent transaction number not reported if lock timeout occurs
DESCRIPTION:
NOTES:
[08-aug-2018] ::: ACHTUNG :::
Important change has been added in FB 4.0.
According to doc\\README.read_consistency.md, read committed isolation level
was modified and new transaction with RC effectively is launched like:
SET TRANSACTION READ COMMITTED READ CONSISTENCY
Moreover, it is unable to start transaction in NO_record_version until
config parameter ReadConsistency will be changed from 1(default) to 0.
This mean that now it is unable to use NO_record_version setting in RC mode
with default firebird.conf ==> we can not check behaviour of engine exactly
as ticket says in its case-1:
===
set transaction read committed no record_version lock timeout 10;
select * from test;
===
For this reason it was decided to create separate section for major version 4.0
and use UPDATE statement instead of 'select * from test' (UPDATE also must READ
data before changing).
JIRA: CORE-2988
FBTEST: bugs.core_2988
"""
import pytest
from firebird.qa import *
from firebird.driver import TPB, TraAccessMode, Isolation
# version: 3.0
substitutions_1 = [('record not found for user:.*', ''),
('-concurrent transaction number is.*', '-concurrent transaction number is'),
('-At block line: [\\d]+, col: [\\d]+', '-At block line')]
init_script_1 = """"""
db_1 = db_factory(sql_dialect=3, init=init_script_1)
test_script_1 = """
drop user tmp$2988;
commit;
create user tmp$2988 password 'tmp$2988';
commit;
recreate table test (id integer);
insert into test values(1);
commit;
revoke all on all from tmp$2988;
grant select,update on test to tmp$2988;
commit;
set transaction lock timeout 1;
update test set id = id;
set term ^;
execute block as
declare v_usr char(31) = 'tmp$2988';
declare v_pwd varchar(20) = 'tmp$2988';
declare v_connect varchar(255);
declare v_dummy int;
begin
execute statement ('update test set id = -1')
with autonomous transaction
as user v_usr password v_pwd
into v_dummy;
end
^
set term ;^
rollback;
set transaction read committed no record_version lock timeout 1;
update test set id = id;
set term ^;
execute block as
declare v_usr char(31) = 'tmp$2988';
declare v_pwd varchar(20) = 'tmp$2988';
declare v_connect varchar(255);
declare v_dummy int;
begin
execute statement ('select id from test')
with autonomous transaction
as user v_usr password v_pwd
into v_dummy;
end
^
set term ;^
rollback;
set list on;
set transaction read committed no record_version lock timeout 1;
select id from test with lock;
set term ^;
execute block as
declare v_usr char(31) = 'tmp$2988';
declare v_pwd varchar(20) = 'tmp$2988';
declare v_connect varchar(255);
declare v_dummy int;
begin
begin
v_dummy = rdb$get_context('SYSTEM', 'EXT_CONN_POOL_SIZE');
rdb$set_context('USER_SESSION', 'EXT_CONN_POOL_SUPPORT','1');
when any do
begin
end
end
execute statement ('select id from test with lock')
with autonomous transaction
as user v_usr password v_pwd
into v_dummy;
end
^
drop user tmp$2988
^
commit
^
-- ||||||||||||||||||||||||||||
-- ###################################||| HQBird 3.x SS/SC |||##############################
-- ||||||||||||||||||||||||||||
-- If we check SS or SC and ExtConnPoolLifeTime > 0 (avaliable in HQbird 3.x) then current
-- DB (bugs.core_NNNN.fdb) will be 'captured' by firebird.exe process and fbt_run utility
-- will not able to drop this database at the final point of test.
-- Moreover, DB file will be hold until all activity in firebird.exe completed and AFTER this
-- we have to wait for <ExtConnPoolLifeTime> seconds after it (discussion and small test see
-- in the letter to hvlad and dimitr 13.10.2019 11:10).
-- This means that one need to kill all connections to prevent from exception on cleanup phase:
-- SQLCODE: -901 / lock time-out on wait transaction / object <this_test_DB> is in use
-- #############################################################################################
execute block as
begin
if ( rdb$get_context('USER_SESSION', 'EXT_CONN_POOL_SUPPORT') = '1' ) then
begin
-- HQbird is tested now:
-- execute statement 'delete from mon$attachments where mon$attachment_id != current_connection';
execute statement 'ALTER EXTERNAL CONNECTIONS POOL CLEAR ALL';
end
end
^
commit
^
"""
act_1 = isql_act('db_1', test_script_1, substitutions=substitutions_1)
expected_stdout_1 = """
ID 1
"""
expected_stderr_1 = """
Statement failed, SQLSTATE = HY000
record not found for user: TMP$2988
Warning: ALL on ALL is not granted to TMP$2988.
Statement failed, SQLSTATE = 40001
deadlock
-update conflicts with concurrent update
-concurrent transaction number is 18
-At block line: 7, col: 9
Statement failed, SQLSTATE = 40001
deadlock
-read conflicts with concurrent update
-concurrent transaction number is 21
-At block line: 7, col: 9
Statement failed, SQLSTATE = 40001
deadlock
-read conflicts with concurrent update
-concurrent transaction number is 24
-At block line: 7, col: 9
"""
@pytest.mark.version('>=3.0,<4.0')
def test_1(act_1: Action):
act_1.expected_stdout = expected_stdout_1
act_1.expected_stderr = expected_stderr_1
act_1.execute()
assert (act_1.clean_stderr == act_1.clean_expected_stderr and
act_1.clean_stdout == act_1.clean_expected_stdout)
# version: 4.0
substitutions_2 = [('^((?!concurrent transaction number is).)*$', ''),
('[\\-]{0,1}concurrent transaction number is [0-9]+', 'concurrent transaction number is')]
init_script_2 = """
create table test(id int, x int, constraint test_pk primary key(id) using index test_pk);
commit;
insert into test(id, x) values(1, 111);
commit;
"""
db_2 = db_factory(sql_dialect=3, init=init_script_2)
act_2 = python_act('db_2', substitutions=substitutions_2)
expected_stdout_2 = """
concurrent transaction number is 13
concurrent transaction number is 13
"""
@pytest.mark.version('>=4.0')
def test_2(act_2: Action, capsys):
with act_2.db.connect() as con:
tx1 = con.transaction_manager()
tx1.begin()
cur1 = tx1.cursor()
cur1.execute('update test set x = ? where id = ?', (222, 1))
# **INSTEAD** of ticket case-1:
# set transaction read committed no record_version lock timeout N;
# -- we start Tx with lock_timeout using custom TPB and try just to **update** record which is locked now
# (but NOT 'SELECT ...'! It is useless with default value of confign parameter ReadConsistency = 1).
# Message about concurrent transaction (which holds lock) in any case must appear in exception text.
# NB: NO_rec_version is USELESS in default FB 4.0 config!
custom_tpb = TPB(access_mode=TraAccessMode.WRITE,
isolation=Isolation.READ_COMMITTED_RECORD_VERSION,
lock_timeout=1)
tx2 = con.transaction_manager(default_tpb=custom_tpb.get_buffer())
tx2.begin()
cur2 = tx2.cursor()
try:
cur2.execute('update test set x = ? where id = ?', (333, 1))
except Exception as e:
print('Exception in cur2:')
print('-' * 30)
for x in e.args:
print(x)
print('-' * 30)
finally:
tx2.commit()
# This is for ticket case-2:
# set transaction read committed lock timeout N;
# select * from test with lock;
custom_tpb.isolation = Isolation.CONCURRENCY
tx3 = con.transaction_manager(default_tpb=custom_tpb.get_buffer())
tx3.begin()
cur3 = tx3.cursor()
try:
cur3.execute('select x from test where id = ? with lock', (1,))
for r in cur3:
print(r[0])
except Exception as e:
print('Exception in cur3:')
print('-' * 30)
for x in e.args:
print(x)
print('-' * 30)
finally:
tx3.commit()
tx1.commit()
act_2.expected_stdout = expected_stdout_2
act_2.stdout = capsys.readouterr().out
assert act_2.clean_stdout == act_2.clean_expected_stdout