Hi, I don't feel strongly one way or another. I do see now that it would be nice to add a section to the Release Notes that describes the API changes (if any), perhaps grouped into some categories.
-Marshall On 6/24/2015 8:19 AM, Richard Eckart de Castilho wrote: > I understand that the "warning" is actually treated like an "error" by the > plugin > in the sense that the plugin exits with a failure state - badly chosen log > level > in my opinion. > > I think the infos are just what the tool found. > > The idea appears to be that whenever there is a change in public API, the > middle > version digit is supposed to change. I did personally also find that a bit > strong. > My feeling was that I only would like to increase the middle version if there > were > significant compatible improvements and for smallish extensions like the ones > we > see below, I'd probably have gone for the last digit. > > However, my gut feeling is that it may be better to stick to the strict rules > of > the semantic versioning and just consider the version to be something that is > basically > managed automatically instead of letting "marketing" influence the versioning > scheme. > > So here I'd say that if we had wanted to do a 2.7.1 release, then we should > have > sticked to changes that do not affect the API in any way. And we didn't just > restore > a dropped API here, right? There are four new methods added and only one of > them > (CasCopier one) was actually a restore, wasn't it? > > Anyway, there are plenty of numbers. We'll hardly run out of them. > > -- Richard > > On 24.06.2015, at 14:04, Marshall Schor <[email protected]> wrote: > >> In this list, I'm guessing the thing that triggered the failure is the >> [WARNING], not the [INFO] - does that seem right? >> >> In that case, this may be an exception, since the [WARNING] change was in >> fact a >> fix restoring the dropped API. >> >> I'm kind of on the fence on this one - though it does seem to me that a fix >> *restoring* a dropped API might semantically feel more like a 2.7.1 than a >> 2.8.0. >> >> Other opinions? >> >> -Marshall >> >> On 6/24/2015 2:49 AM, Richard Eckart de Castilho wrote: >>> Hi all, >>> >>> I changed the configuration of the semantic versioning plugin for >>> uimaj-core yesterday. >>> According to the plugin, the next version should be 2.8.0 because public >>> API has changed: >>> >>> [INFO] >>> --- maven-enforcer-plugin:1.3.1:enforce (enforce-compatibility) @ >>> uimaj-core --- >>> >>> [INFO] Version specified as <2.7.0> >>> [INFO] Using >>> </home/jenkins/jenkins-slave/maven-repositories/0/org/apache/uima/uimaj-core/2.7.0/uimaj-core-2.7.0.jar> >>> as previous JAR >>> [INFO] Using >>> </home/jenkins/jenkins-slave/workspace/UIMA-SDK/trunk/uimaj-core/target/uima-core.jar> >>> as current JAR >>> Class org.apache.uima.jcas.JCas >>> Added Method getAnnotationIndex, sig >>> <T:Lorg/apache/uima/jcas/tcas/Annotation;>(Ljava/lang/Class<TT;>;)Lorg/apache/uima/cas/text/AnnotationIndex<TT;>;, >>> desc (Ljava/lang/Class;)Lorg/apache/uima/cas/text/AnnotationIndex;, access >>> abstract public >>> Added Method getAllIndexedFS, sig >>> <T:Lorg/apache/uima/jcas/cas/TOP;>(Ljava/lang/Class<TT;>;)Lorg/apache/uima/cas/FSIterator<TT;>;, >>> desc (Ljava/lang/Class;)Lorg/apache/uima/cas/FSIterator;, access abstract >>> public >>> Added Method getIndex, sig >>> <T:Lorg/apache/uima/jcas/cas/TOP;>(Ljava/lang/String;Ljava/lang/Class<TT;>;)Lorg/apache/uima/cas/FSIndex<TT;>;, >>> desc (Ljava/lang/String;Ljava/lang/Class;)Lorg/apache/uima/cas/FSIndex;, >>> access abstract public >>> Class org.apache.uima.util.CasCopier >>> Added Method alreadyCopied, desc (Lorg/apache/uima/cas/FeatureStructure;)Z, >>> access public >>> >>> [WARNING] Rule 0: org.semver.enforcer.RequireBackwardCompatibility failed >>> with message: >>> >>> Current codebase is not backward compatible >>> (BACKWARD_COMPATIBLE_IMPLEMENTER) with version <2.7.0>. Compatibility type >>> has been detected as <BACKWARD_COMPATIBLE_USER> >>> >>> >>> So since we have +1s recently for semantic versioning, how about making it >>> a 2.8.0? >>> >>> >>> As a side-note: this broke the build and somebody immediately opened a >>> ticket on >>> Stackoverflow asking about the broken build. Anyway, I don't regret >>> breaking it. >>> IMHO one function of a CI system is to take the build load and waiting time >>> off >>> from the local developer (machines). >>> >>> http://stackoverflow.com/questions/31016180/failed-to-execute-maven-enforcer-plugin/31019408#31019408 >>> >>> Cheers, >>> >>> -- Richard >>> >>> On 23.06.2015, at 16:30, Marshall Schor <[email protected]> wrote: >>> >>>> 2.7.1sdk has 33 issues, including a couple relating to some regressions >>>> around >>>> some CasCopier functionality that was accidentally dropped. >>>> >>>> I'm thinking of releasing it after cleaning up the 2-3 remaining >>>> non-resolved >>>> issues. Any other thoughts pro/con? >>>> >>>> -Marshall >
