I see. Seems that some new warnings have the same properties and are therefore marked as duplicates.

Hello, Ulli.
What kind of properties are you talking about?

Same type, filename, line, ... Currently all warnings are collected in a set which does not contain duplicates.

Code changed in jenkins
User: Ulli Hafner
Path:
src/main/java/hudson/plugins/analysis/core/FilesParser.java
src/main/java/hudson/plugins/analysis/core/ParserResult.java
http://jenkins-ci.org/commit/analysis-core-plugin/1d1eb5b17cc4fa21043fe8332599b66ab1e17334
Log:
Merge pull request #21 from sschuberth/master

Fix checking for duplicate annotations. (Maybe related to JENKINS-24873).

Compare: https://github.com/jenkinsci/analysis-core-plugin/compare/a5cae3ab9dfd...1d1eb5b17cc4

Do you have some warnings in the same line and column?

Yes, I have.

Code changed in jenkins
User: Ulli Hafner
Path:
src/main/java/hudson/plugins/analysis/core/BuildResultEvaluator.java
http://jenkins-ci.org/commit/analysis-core-plugin/c873a76a22edd958c33578713fc317eec5a3df0e
Log:
JENKINS-24873 Fixed priority naming.

Compare: https://github.com/jenkinsci/analysis-core-plugin/compare/b75554ea6be6...c873a76a22ed

Can you please check if this works with the pull request that is part of new analysis-core 1.61 release?

I've installed analysis-core v. 1.61, but nothing was changed.

Did you restart the build? Previous results are persisted and cannot not be changed.

Actually, I've created a new job.

Can you please show the log messages from the PMD plugin at the end of the console log of your job? (And did the message in the job change from "warnings of priority Normal Priority" to "warnings of priority Normal"?)

Can you please show the log messages from the PMD plugin at the end of the console log of your job?

See 1, 2 and 3 for the same test cases correspondingly.

And did the message in the job change from "warnings of priority Normal Priority" to "warnings of priority Normal"?

Yes, it did.
So, the plugin was really installed

Sorry, seems that I did not read the bug report carefully. The described behavior is unrelated to the changes of the pull request.

Currently the algorithm to detect new warnings is not rock-solid. But I think I can improve the algorithm somewhat so that this issue will be fixed. (This is not a general solution, here we need to wait for JENKINS-6675.)

Code changed in jenkins
User: Ulli Hafner
Path:
src/main/java/hudson/plugins/pmd/parser/PmdParser.java
http://jenkins-ci.org/commit/pmd-plugin/c6093bba7feef7a46e56bbb2027db868c763f98d
Log:
[FIXED JENKINS-24873] Use warning type in hash code computation.

Code changed in jenkins
User: Ulli Hafner
Path:
.idea/inspectionProfiles/Project_Default.xml
http://jenkins-ci.org/commit/analysis-suite-plugin/0a6e1b6982b33012fa505f3e91619aebdbb904b7
Log:
[FIXED JENKINS-24873] Use warning type in hash code computation.

Code changed in jenkins
User: Ulli Hafner
Path:
src/main/java/hudson/plugins/dry/parser/DuplicationParserRegistry.java
http://jenkins-ci.org/commit/dry-plugin/51a6950bad1d6684a3dac0a11841cc1b017d91ea
Log:
[FIXED JENKINS-24873] Use warning type in hash code computation.

Code changed in jenkins
User: Ulli Hafner
Path:
src/main/java/hudson/plugins/checkstyle/parser/CheckStyleParser.java
http://jenkins-ci.org/commit/checkstyle-plugin/dff75c9214f32e286c014cb27ab88923bf88c056
Log:
[FIXED JENKINS-24873] Use warning type in hash code computation.

But I think I can improve the algorithm somewhat so that this issue will be fixed.

Ulli, how can I test this fix?

Sorry, didn't find time to make a new release yet. You can download the latest plugin from our CI server, there is a link in the top of the plugin wiki pages.

Thank you for your work, Ulli.

Hello, Ulli.

I've faced with related problem on another project.
Test case:

1. Start the first build.
   See summary result.
2. Change source code, check "Use delta for new warnings" check box and start the second build.
   The numbers of new NORMAL priority warnings are different: 2 in sum result and 3 on new warning page.
3. Uncheck "Use delta for new warnings" check box and start the third build.
   The numbers of new NORMAL priority warnings become the same: 3 in sum result and 3 on new warning page.

It seems that is case "Use delta for new warnings" check box is checked, another algorithm is used only for build summary page. But for a new warnings page generation is used the same algorithm as in case "Use delta for new warnings" check box is unchecked.

I suppose that number of new warnings of a particular priority on new warning page should always be the same as on build summary page.

Any updates, Ulli?

Actually no, I'm not sure what I can do here. The delta algorithm cannot be used to compute the actual new warnings, since it only subtracts the totals of two builds. Maybe I should remove the flag, seems only to be helpful if the other algorithm is too slow due to a lot of warnings.

Do you mean that usage of different algorithms on a build summary page and new warnings page while "Use delta for new warnings" is checked is expected plugin behaviour?

Well, yes

Actually, we use delta checkbox because the other algorithm too frequently finds false positive new warnings. But, as I've got, this should be fixed in JENKINS-6675.

One of my students is currently improving the new warning detection using a new algorithm that is based on the Java syntax tree. Hopefully, this will reduce the number of false positives. However, you can easily find an example of a false positive for every algorithm...

Hi Ulli, after updating we now have the warning that all warnings are new (and fixed at the same time). May this problem relates to this issue?