This should be fixed now. There's a knob in Jenkins to switch off " Incremental build - only build changed modules" to always build all modules. Thanks Olivier for the pointer (http://mail-archives.apache.org/mod_mbox/www-builds/201211.mbox/%3CCAPoyBqS9i9AxVAMyzTZeLp3xB%2Bn1Oo0ts8O1x9P-yQ%3DSXY5Lvw%40mail.gmail.com%3E).
--Pei > -----Original Message----- > From: Chen, Pei [mailto:[email protected]] > Sent: Wednesday, November 28, 2012 4:18 PM > To: [email protected] > Subject: RE: spurious build errors from Jenkins > > According to the build logs, it looks like it is only building only the > module that > changed from SCM rather than all modules every time. > Does anyone know if there's a knob to switch and if so, which one to set this > in Jenkins? We can take a closer look once we get access to modify the jobs > again... > > For example: 64 when the only change was to the constituency parser: > [INFO] > ------------------------------------------------------------------------ > [INFO] Reactor Build Order: > [INFO] > [INFO] Apache cTAKES Constituency Parser [INFO] Apache cTAKES > CoReference Resolver [INFO] Apache cTAKES Relation Extractor [INFO] > Apache cTAKES Distribution > > Vs: https://builds.apache.org/job/cTAKES-trunk-compiletest/65/consoleFull > when invoked by clicking on 'build now': > [INFO] Reactor Build Order: > [INFO] > [INFO] Apache cTAKES > [INFO] Apache cTAKES common type system > [INFO] Apache cTAKES utils > [INFO] Apache cTAKES core > [INFO] Apache cTAKES part-of-speech tagger [INFO] Apache cTAKES chunker > [INFO] Apache cTAKES document preprocessor [INFO] Apache cTAKES > dictionary lookup [INFO] Apache cTAKES context dependent tokenizer > [INFO] Apache cTAKES LVG lexical tools [INFO] Apache cTAKES named entity > contexts [INFO] Apache cTAKES Constituency Parser [INFO] Apache cTAKES > CoReference Resolver [INFO] Apache cTAKES Drug NER [INFO] Apache > cTAKES Side Effects [INFO] Apache cTAKES Smoking Status [INFO] Apache > cTAKES Dependency Parser [INFO] Apache cTAKES Assertion [INFO] Apache > cTAKES ctakes-clinical-pipeline [INFO] Apache cTAKES Relation Extractor > [INFO] Apache cTAKES Pad Term Spotter [INFO] Apache cTAKES Temporal > Information Extraction [INFO] Apache cTAKES Distribution > > > -----Original Message----- > > From: Steven Bethard [mailto:[email protected]] > > Sent: Friday, November 16, 2012 4:17 PM > > To: [email protected] > > Subject: spurious build errors from Jenkins > > > > So as you may have noticed, we just experienced the same kind of > > spurious build error with ctakes-relation-extractor that we were > > seeing with ctakes- assertion before. We got errors like: > > > > Could not resolve dependencies for project > > org.apache.ctakes:ctakes-core:jar:3.1.0-incubating-SNAPSHOT: > > The following artifacts could not be resolved: > > org.apache.ctakes:ctakes-type-system:jar:3.1.0-incubating-SNAPSHOT, > > org.apache.ctakes:ctakes-utils:jar:3.1.0-incubating-SNAPSHOT > > > > This is exactly the error that we get if you run `mvn compile` from > > the `ctakes-core` directory, which we should never be doing. And > > there's no error if we run `mvn compile` from the top-level directory, > > which is what we should always be doing. > > > > So I suspect that Jenkins is somehow configured to run `mvn compile` > > separately for the separate modules. Can we turn that off, and only > > run the `mvn compile` from the top level? There should be no need to > > run it for each module if we run it from the top. > > > > > > Steve > > > > P.S. To make Jenkins happy again, all you have to do is start a build > > manually with the "Build Now" button in Jenkins. But we shouldn't need to > do that.
