2016-02-19
| ||
18:55 | • Ticket [dc9b1c91] One trigger may use auxdata belonging to another. status still Closed with 3 other changes (artifact: c1b47ba6 user: dan) | |
18:55 | • Closed ticket [dc9b1c91]. (artifact: 0395b295 user: dan) | |
18:55 | • Ticket [dc9b1c91]: 6 changes (artifact: 0bea5f1c user: dan) | |
18:54 | Use a separate list of aux-data structures for each trigger program at the VDBE level. Fix for [dc9b1c91]. (check-in: c4295725 user: dan tags: trunk) | |
17:22 | • New ticket [dc9b1c91] One trigger may use auxdata belonging to another.. (artifact: 018bf9a3 user: dan) | |
Ticket Hash: | dc9b1c91db15ae9d76e98ee7a6d8125613307085 | |||
Title: | One trigger may use auxdata belonging to another. | |||
Status: | Closed | Type: | Code_Defect | |
Severity: | Critical | Priority: | Immediate | |
Subsystem: | Unknown | Resolution: | Fixed | |
Last Modified: | 2016-02-19 18:55:55 | |||
Version Found In: | ||||
User Comments: | ||||
dan added on 2016-02-19 17:22:13:
Assuming the REGEXP operator implementation uses sqlite3_set_auxdata() to store a compiled version of its regular expression (as the bundled extension does), test 1.1 of the following tcl test script fails: do_execsql_test 1.0 { CREATE TABLE t1(a, b, c); CREATE TABLE x1(x, y, z); CREATE TABLE x2(x, y, z); CREATE TRIGGER tr1 AFTER INSERT ON t1 BEGIN INSERT INTO x1 VALUES( new.a REGEXP 'abc', new.b REGEXP 'abc', new.c REGEXP 'abc' ); END; CREATE TRIGGER tr2 AFTER INSERT ON t1 BEGIN INSERT INTO x2 VALUES( new.a REGEXP 'def', new.b REGEXP 'def', new.c REGEXP 'def' ); END; INSERT INTO t1 VALUES('abc', 'def', 'abc'); SELECT * FROM t1; } {abc def abc} do_execsql_test 1.1 { SELECT * FROM x1 } {1 0 1} do_execsql_test 1.2 { SELECT * FROM x2 } {0 1 0} The trouble is that the REGEXP operators in trigger "tr1" are incorrectly using the compiled regular expressions for trigger "tr2" - causing the operators in "tr1" to populate table x1 with the same values as table x2. If the two triggers were using different operators or user-functions that both used sqlite3_set_auxdata() a crash or other error might have occurred instead. Reported on mailing list here: http://osdir.com/ml/general/2016-02/msg26665.html dan added on 2016-02-19 18:55:28: Fixed by [c4295725]. |