... to not dissipate and reduce our (effort) losses :) Howdy,
In short, I think we can agree on this sentence: "we need to get rid of maven 3.8.x as fast as possible, release maven 3.9.x and once out, keep it on regression fixes until maven 4 is ready". # Get rid of Maven 3.8.x Hence, I think we MUST do 3.8.6, as there is one regression that it IMHO a must: https://issues.apache.org/jira/issues/?jql=project%20%3D%20MNG%20AND%20resolution%20%3D%20Unresolved%20AND%20fixVersion%20%3D%203.8.6%20ORDER%20BY%20priority%20DESC%2C%20updated%20DESC There are some other issues "affects 3.8.5" but I don't see them as blockers (but please argue if you disagree): https://issues.apache.org/jira/issues/?jql=project%20%3D%20MNG%20AND%20resolution%20%3D%20Unresolved%20AND%20affectedVersion%20%3D%203.8.5%20ORDER%20BY%20priority%20DESC%2C%20updated%20DESC Out of them: MNG-7432 [REGRESSION] Dependencies from profile not picked up via -P<profileName> Must be fixed (PR merge pending) MNG-7433 [REGRESSION] Multiple maven instances working on same source tree can lock each other Fluke IMHO, but as Dan is an old Maven biker, I think he uses Takari Lifecycle plus maybe even Takari Local Repo extension, unsure. Definitely this looks too vague to me, as Guillaume PR for sure cannot deadlock two JVMs.... (it uses in-JVM locks). MNG-7449 [REGRESSION] StringVisitorModelInterpolator NullPointerException Anyone? Unsure what is happening here, also how are these "new classes in 3.8.5"? MNG-7441 Update Version of Logback to Address CVE-2021-42550 We should do this, as we plan to keep 3.8.x on the shelf for a while. Hence, we will get more and more reports like these from end users. Let's cut it from the root. MNG-7438 add execution id to "Configuring mojo xxx with basic configurator" debug message IMHO: nope MNG-4917 Profile not active even though it has activeByDefault set to true IMHO: nope -- this issue is 12 years old! MNG-7429 The classloader containing build extensions should be used throughout the build IMHO nope: (unless regression from some 3.8.x) MNG-7448 if we delete bin directory in apache-maven/src, we can not find because git has ignore it . IMHO nope: ??? MNG-7439 Make fields of CliRequest protected IMHO nope So, let's agree on what to fix for 3.8.6 and push it out... and once 3.8.6 out, forget it, no more fixes (unless some really serious regression is reported, but otherwise nada). # Maven 3.9.x IMHO, Maven 3.9.0 should wait for resolver 1.8.x (has one remaining PR ready to merge) as then Maven will get much much more goodness (not only 1.7.x locking, but also extensible checksums, smart checksums, provided checksums, DF and BF collector, ability for signature resolution, proper ignores for artifacts not needed checksums). To make smart checksums usable for maven, http transport is needed as well (PR as it is no go, it needs to make Wagon still remain default transport, will fix that). Or, we could make it as is, and then if anyone has issues or badly wants Wagon, there is a command line to make it so.... So, IMHO, for 3.9.x - Port fixes that we decide to be done in 3.8.6 naturally - Use resolver 1.8.0 (wait for it, one remaining PR merge, released hopefully very soon) - Apply https://github.com/apache/maven/pull/635 and decide: keep Wagon default or not. Also this PR would get rid of shaded stuff as well. And that's pretty much it. Keep 3.9.x on regression fixes only (so only if some serious regression falls in). # Maven 4 Guillaume already did a huge amount of work that he probably needs help on, so help him to wrap up things.... WDYT? T
