Cluster segfault when invalid LUA is evaluated #1105
Description
This one is easy to reproduce:
- Create a test with the following code:
public function testEvalWithSyntaxError() {
$scr = "return KEYS[1,]";
$result = $this->redis->eval($scr, Array(0), 1);
$this->assertFalse($result);
}- Run cluster tests
At some point (it seems random), the process will segfault with the following trace:
* thread #1: tid = 0x0000, 0x000000010afec631 php`_emalloc + 204, stop reason = signal SIGSTOP
* frame #0: 0x000000010afec631 php`_emalloc + 204
frame #1: 0x000000010b011810 php`add_assoc_string_ex + 54
frame #2: 0x000000010c125c83 redis.so`redis_parse_info_response(response=<unavailable>, z_ret=<unavailable>) + 275 at library.c:970 [opt]
frame #3: 0x000000010c142bba redis.so`cluster_info_resp(execute_data=<unavailable>, return_value=0x000000010c217ad0, c=0x0000000112345000, ctx=<unavailable>) + 58 at cluster_library.c:2034 [opt]
frame #4: 0x000000010c13ba6e redis.so`zim_RedisCluster_info(execute_data=0x000000010c217b30, return_value=0x000000010c217ad0) + 382 at redis_cluster.c:2755 [opt]
frame #5: 0x000000010b060c4c php`ZEND_DO_FCALL_SPEC_RETVAL_USED_HANDLER + 667
frame #6: 0x000000010b04b36f php`execute_ex + 44
frame #7: 0x000000010b04b5e7 php`zend_execute + 551
frame #8: 0x000000010b00de96 php`zend_execute_scripts + 307
frame #9: 0x000000010afb59e2 php`php_execute_script + 770
frame #10: 0x000000010b0aa11f php`do_cli + 3806
frame #11: 0x000000010b0a90b8 php`main + 1249
frame #12: 0x00007fff979665ad libdyld.dylib`start + 1
It seems to be some issue with the INFO parse function, but my knowledge of this library is limited and can't work on a fix.