Page MenuHomePhabricator

CirrusSearch blows up on really really long queries in production
Closed, ResolvedPublic

Description

CirrusSearch fails on really really long queries like this:
asdfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffasdfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffasdfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffasdfffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffffff


Version: unspecified
Severity: normal

Details

Reference
bz55798

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 22 2014, 2:40 AM
bzimport added a project: CirrusSearch.
bzimport set Reference to bz55798.
bzimport added a subscriber: Unknown Object (MLST).

But this doesn't seem to happen in dev.

This is caused by pcre backtracking - in prod we must have the limit lower than in dev. Either way, I'll remove the offending regex and replace it with a loop over the characters. It'll probably be faster any way.

Change 90200 had a related patch set uploaded by Manybubbles:
Improvements when searching long strings

https://gerrit.wikimedia.org/r/90200

Change 90202 had a related patch set uploaded by Manybubbles:
Tests for unbalanced quotes

https://gerrit.wikimedia.org/r/90202

Change 90200 merged by jenkins-bot:
Improvements when searching long strings

https://gerrit.wikimedia.org/r/90200

Change 90202 merged by jenkins-bot:
Tests for unbalanced quotes

https://gerrit.wikimedia.org/r/90202