-
Bug
-
Resolution: Fixed
-
Major
-
None
Similar issue to JENKINS-7535, but in the Ivy Plugin instead of the Maven Plugin.
Every time a project is saved, Jenkins rebuilds the dependency graph.
For Ivy Projects this includes building up a dependency map of all the Ivy modules in Jenkins and then applying the build triggers based on that information. Currently, this is being done in every single module resulting in an operation which is roughly in the order of O(n^2 + xn) where n is the number of Ivy modules in Jenkins and x is the average number of direct dependencies each module has.
[JENKINS-12638] Rebuilding dependency graph slow on large numbers of Ivy Projects
Code changed in jenkins
User: tbingaman
Path:
src/main/java/hudson/ivy/IvyModule.java
http://jenkins-ci.org/commit/ivy-plugin/cead62a6ab51f30bb24f9e62dc2febdb70778950
Log:
[FIXED JENKINS-12638] Make dependency graph calculation more efficient by storing intermediate computational data with the DependencyGraph.
Integrated in 
plugins_ivy #61
[FIXED JENKINS-12638] Make dependency graph calculation more efficient by storing intermediate computational data with the DependencyGraph. (Revision cead62a6ab51f30bb24f9e62dc2febdb70778950)
Result = SUCCESS
tbingaman :
Files :
- src/main/java/hudson/ivy/IvyModule.java
The second part of that (O(xn)) can't be helped, but the first part (O(x^2)) is horribly inefficient.
Using the solution that was applied to the Maven Plugin for
JENKINS-7535in commit 6a5330fd936fb949da74d738b82dcf5333282b61 this first part can be reduced roughly to an O(n) operation, resulting in a final order of about O(n + xn).