Hervé Boutemy

in the reproducible-central page, there are link to diffoscope results that show the different problems found, for example https://github.com/jvm-repo-rebuild/reproducible-central/blob/master/content/org/apache/activemq/activemq-parent-5.17.0.diffoscope

to me, once this PR is merged, the only non-reproducible bits are a stupid timestamp, I'll need to chase in Spring later ``` ├── META-INF/spring.handlers │ @@ -1,3 +1,3 @@...

Using Reproducible Build timestamp is just a choice: it's easy to update the plugin to detect and use it. But if UUID is not intended to be reproducible, using reproducible...

it's a choice I respect: I switched myself from putting timestamps everywhere for tracking every build result to remove everything with Reproducible Builds, I know that each approach has pros...

perhaps adding an option for users to choose if they want this timestamp or not is a better choice (like JAXB that provides `-no-header` option for such case)

yes, we do it with maven-replacer-plugin I just added the workaround for ANTLR4 https://github.com/antlr/antlr4/pull/3809/commits/270f13b12f779e371e431eac08544feb7ad9fd7c

one question: there are 2 sources of non-reproducible bit 1. the timestamp; that we can postprocess with maven-replacer-plugin 2. the AST rewrite elements, like ``` // AST REWRITE - //...

@elharo is this ok now to merge, or do you think something additional is required?

can't merge, there are 4 ITs failing: https://builds.apache.org/blue/organizations/jenkins/maven-box%2Fmaven-archetype/detail/ARCHETYPE-584/14/pipeline :(

@khmarbaise on the second point (automatic site publish during release): this can't work for multi-module builds, because such build require site:stage during initial build then another build to do the...