Can you please check if this also happens after refreshing the page with CTRL-F5? Does the size of the graph change if you modify the parameters?

And is this is a single project in the view or are there several ones?

Multiple projects. It is not a cache problem and is happening on all our Hudson clusters. The size parameters works fine.

As you can see from the screenshots the priority distribution changes vastly but the New vs Fixed does not work at all. Similar issues on all various types of portlets. Also we have the latest versions installed of plugins and core components. The portlets have never worked for us (E.g. it is not a newly introduced issue).

Integrated in Hudson Plug-ins (Compile) #75
[FIXED JENKINS-8283] Use 'today' - 'jobTimestamp' when computing the builds that should be considered in a trend graph.

Since I changed the algorithm in several aspects I hope that now everything works fine. If possible, it would be nice if you can download the plugins from our build server and install them manually and test these before I make a public release: http://faktorzehn.org:8081/job/Hudson%20Plug-ins%20(Compile)/75/

You need to install the analysis-core and the checkstyle plug-in.

Hi! Thanks for the prompt response. However, the result gets even stranger this time around. Please check the attached screenshot (CaptureFromSnapshotVersion.png).

Ok, I see. There are still builds considered that should be out of the specified range. Is the number of warnings in the left side correct?

On the right side: do you have new or resolved warnings in the corresponding builds or is the increasing total number only because of adding new projects?

The warnings are due to changes in the code being built. But please notice that the two graphs are not both the same type of warnings:

• Checkstyle warnings
  and
• Compiler warnings

I wanted just to show that the two portlets renders the dates on the horizontal axis quite differently.

More info: If I change to # 30 days, the same date range is being used, but more dates are being printed in the the time-axis. Really strange.

Thanks again, I think I found the problem. My build time check is after the first value has been added That means that there are always some wrong values for projects that have no build result in the given period.

Do you have some jobs that have no result in the last 4 days?

Integrated in Hudson Plug-ins (Compile) #77
[FIXED JENKINS-8283] Check the build times before adding the first result (since a project might not have any builds in the defined range).

I think I finally fixed it now

One drawback of the fix: if a project has no builds in the selected range, then there a no warnings counted. Maybe I should change that and reuse the last available result for that build, and use that result for all the visible days (as this is done for builds that have no result on a specific day). What do you think?

Reuse the last results sounds fine to me.

And again, thanks for the quick response and fix

And the new version is now tested and seems to work fine. Time for a new release?

I'm waiting for a pending change in the findbugs plugin. Hopefully, there will be a christmas release