-
Improvement
-
Resolution: Fixed
-
Major
-
None
-
Platform: All, OS: All
A couple of weeks after we started using the sloccount plugin our hudson started
to crash due to out of memory exceptions. This seems to happen due to a very
high number of instances of hudson.plugins.sloccount.model.File
It do seem that we get one such instance per file, per build, per job.
Secondly our build.xml files (per build, per job) have grown a lot.
We dont uses / needs the ability to browse the size of individual files.
Would it be possibel to have the sloccount plugin configurable, so only the
SLOCCountTrend graph and the delta information on the build page was available
This corresponds to the amount of information available at the 'Languages' tab
on the 'SLOCCount Results' page
Alternatively the plugin could read the detailed info from the sloccount.sc file
if a user clicked the files or folder tabs. (Lazy reading)
That would ofcourse require that the sloccount.sc was kept in the builds folder
as the violations plugin does with the violations.xml file
- is related to
-
JENKINS-21921 Lazy loaded report details are never released
-
- Closed
-
[JENKINS-4769] Memory consumption is huge
Michal Turek
added a comment - Fixed in https://github.com/mixalturek/sloccount-plugin/commit/284e6ea145aaa458078b6d16972dcc305a742836.
Michal Turek
added a comment - Fixed in https://github.com/mixalturek/sloccount-plugin/commit/284e6ea145aaa458078b6d16972dcc305a742836 . Code changed in jenkins
User: Michal Turek
Path:
src/main/java/hudson/plugins/sloccount/ReportSummary.java
src/main/java/hudson/plugins/sloccount/SloccountBuildAction.java
src/main/java/hudson/plugins/sloccount/SloccountChartBuilder.java
src/main/java/hudson/plugins/sloccount/SloccountPublisher.java
src/main/java/hudson/plugins/sloccount/SloccountResult.java
src/main/java/hudson/plugins/sloccount/model/SloccountLanguageStatistics.java
src/main/java/hudson/plugins/sloccount/model/SloccountParser.java
src/main/java/hudson/plugins/sloccount/model/SloccountPublisherReport.java
src/main/java/hudson/plugins/sloccount/model/SloccountReport.java
src/main/java/hudson/plugins/sloccount/model/SloccountReportInterface.java
http://jenkins-ci.org/commit/sloccount-plugin/284e6ea145aaa458078b6d16972dcc305a742836
Log:
JENKINS-4769 Memory consumption is huge
- New data format introduced, it should be backward compatible.
- Only statistics per language are stored in the publisher.
- The legacy structures are still stored too but with no data (they are null). This is to be able to load the legacy data that were stored using old version of the plugin.
- The report data can loaded from the original files after they are needed.
- Trend graph updated to use the new language statistics.
- Report summary updated to use the new language statistics.
- The plugin now consumes much less memory and is much faster most of the time.
- Only statistics are used during results publication at build time, displaying trend graph and displaying report summary. Only small statistics are deserialized.
- Complete report is loaded only when report details page is displayed.
Code changed in jenkins
User: Michal Turek
Path:
src/main/resources/hudson/plugins/sloccount/SloccountResult/files.jelly
src/main/resources/hudson/plugins/sloccount/SloccountResult/folders.jelly
src/main/resources/hudson/plugins/sloccount/SloccountResult/languages.jelly
http://jenkins-ci.org/commit/sloccount-plugin/e4050c883f0604d14fb3f6dd8ea9bffe119ef858
Log:
JENKINS-4769 Memory consumption is huge
- Report is cached at report details pages.
Code changed in jenkins
User: Michal Turek
Path:
src/main/java/hudson/plugins/sloccount/ReportSummary.java
http://jenkins-ci.org/commit/sloccount-plugin/b7c0907a8c738ec0cc1ed03d2bc3acdb15828f04
Log:
JENKINS-4769 Memory consumption is huge
- Previous commit added a regression, thousands in numbers are no longer separated by commas in report summary.
- Displaying of all numbers wrapped to StringUtil.grouping() to fix the regression.
build.xml is not the place to log the per run full sloccount data. If there is a need, only some summary could be included (as with the compiler warnings). The sloccount data should log its per run to a different xml file which is only consulted when drilling down into a run result.
https://groups.google.com/forum/#!topic/jenkinsci-users/z66arr7QN1M