On Thu, Apr 16, 2020 at 9:52 AM Enrico Olivelli <[email protected]> wrote: > > Christopher, > answers inline. > Thank you for testing ! > > Il giorno gio 16 apr 2020 alle ore 15:12 Christopher <[email protected]> > ha scritto: > > > -0 (non-binding), I did find some problems that prevent some maven > > profiles, including fatjar, from being activated, among other minor > > issues; nothing too serious, but the inability to run the fatjar > > profile might be a blocker for some, as might be the missing patch for > > ipv6 comparisons in the C client. > > > > Good > > * I tested the convenience binary with Apache Accumulo 2.0.0 and basic > > functionality is all there > > * I checked the LICENSE.txt/NOTICE.txt files for expected content > > (copyright year was 2020, as expected) > > * Provided SHA512 signatures and GPG signatures match the tarballs > > (though the .sha512 files don't have terminating EOL chars) > > > > 3cc33e7630eb47e5807bc90610ae084c603960645ab36d4d6f775715ea75a7041835507029a2ca815ae16f4b8110bf9001a602ed78f0a7866c11dc15643b747c > > apache-zookeeper-3.6.1-bin.tar.gz > > > > 21741f5ee09a8ad897da965c4e3570e4dd7d3a24bf990a8d77738144f4ca883ae6ccf86eb6f9a248c772ef2a22eaed438f4f3313166f89b8e28448d59a6ea7bd > > apache-zookeeper-3.6.1.tar.gz > > * The contents of the source tarball match the contents of the tag > > (3ed3a9890472b251f9a6241317feef5f02cc0692) > > * Was able to build from source using `mvn clean verify -Pfull-build > > -DskipTests` > > > > Bad > > * The tag (and source tarball) is missing the commit for '754cf015f > > ZOOKEEPER-3726: invalid ipv6 address comparison in C client', which is > > present in the branch-3.6 branch > > ** This might be resolvable by simply updating the JIRA to mark > > 3.6.2 as the fixVersion instead of 3.6.1, unless it's critical to > > include > > > > Done. If ZOOKEEPER-3726 is not a blocker then I feel we can live without it. > We can add it in case of the need of a new iteration.
Agreed. Makes sense. > > > * Saw a few unit test failures on Fedora 31 x86_64 with Maven 3.6.3 > > and java-13-openjdk-13.0.2.8-1.rolling.fc31.x86_64 (I ran `mvn clean > > package`) > > ** org.apache.zookeeper.server.util.RequestPathMetricsCollectorTest > > failed with AssertionError > > ** org.apache.zookeeper.server.quorum.QuorumDigestTest failed with > > AssertionError > > > Does it pass if you rerun it again ? This test is not failing for me (on > Linux + jdk8) They both pass on a second run, when I ran them in isolation. It might just be because surefire forkCount is 8, and my laptop is slow. Not sure. Would need further investigation. I'm not worried about this, though, and wouldn't consider it a blocker... but I might open up a JIRA if I see it again and can capture a stack trace or logs. > > > > * release did not appear to be prepared using the maven-release-plugin > > from the branch-3.6, but from a different (local?) branch; this > > resulted in a few minor issues > > > > yes, the tradition here is to create a work branch release-3.6.1 and then > it up to the Release Manager to handle the status of that branch > it is not strictly the Maven way, but we discussed that approach while > releasing 3.6.0, that was the first release with the maven-release-plugin > > this is our guide > https://cwiki.apache.org/confluence/display/ZOOKEEPER/HowToRelease+using+maven+release+plugin > I recommend using the <pushChanges>false</pushChanges> and <localCheckout>true</localCheckout> in maven-release-plugin (for release:prepare and release:perform) and using '3.6.1' instead as the release version. Improving the build and release process is something I have a lot of experience with, so I may look into this more later. Much of the details on this page can be replaced with an interactive helper script, as I've done for both Accumulo and Fluo already. > > > > ** git commit hashes don't match up with the branch (commits appear > > cherry-picked and exclude ZOOKEEPER-3726, mentioned above) > > > This is not a big deal, it is important that we have the good tag > (release-3.6.1-0) Agreed. This is minor. It was just confusing. Missing commits in the release can be included in future releases. A bigger problem would be if the commits only existed in the tag and not in the maintenance branch, because that might mean a regression in the next release. The main concern I had here was that the JIRA issue was marked as fixed for this release, but it wasn't included. > > ** the content of the pom.xml's <scm> <tag> includes `-0`, which > > will not be the final tag name if the artifacts are approved for > > release (actual tag should be "release-3.6.1") > > > > I agree, this is not nice. We can improve it. > Not a blocker for this release > > > > * fatjar profile is broken because fatjar module and zookeeper-it > > module specify wrong parent pom version (bad cherry-pick from > > master/3.7.0-SNAPSHOT?) > > > > This may be a problem that leads to the inability of building the source > release. > If you do not have ever built 3.7.0-SNAPSHOT locally you won't be able to > build from the released sources. > The source tarball we release is actually the main (an only) release > artifact and it should be buildable. > > I will double check, but I feel this can be a showstopper for this RC > > Christopher, do you want to send a fix patch for branch-3.6 ? I think the core of the problem is that not all modules are activated during `mvn release:prepare` when the POM versions are updated. However, you can't activate all modules at once because `fatjar` and `full-build` profiles are mutually exclusive. This is something I've already fixed in my PR for ZOOKEEPER-3791. I didn't anticipate backporting my PR for ZOOKEEPER-3791 to 3.6, but... the work was already done, and I'm pretty sure it fixes the issue here. So, I added an additional commit to that PR that updates the `<preparationGoals>` to activate all modules (and to use `verify` instead of `install`, since `install` is generally not advised... and can specifically cause problems with `release:prepare` creating different local artifacts than what are staged in `release:perform`). Backporting my PR to 3.6 should fix the problem, but you should test it. > > Enrico > > > > > > Did not check > > * did not run any unit tests from modules other than zookeeper-server, > > since the two test failures mentioned above terminated the build > > prematurely, and I didn't feel like running it again to skip those. :) > > > > You can re-run the build. Or if you want you can build the full repository > with -DskipTests and then build with the "-rf" option (Resume from) and > start the tests from the module after zookeeper-server. > Actually we still have some flaky tests that should be improved, but this > is not a blocker for a release. > > Enrico > > > > > > > > On Thu, Apr 16, 2020 at 6:34 AM Szalay-Bekő Máté > > <[email protected]> wrote: > > > > > > +1 (non-binding) > > > > > > - I built the source code on Ubuntu 18.4 using OpenJDK 8u242 and maven > > > 3.6.3. > > > - All the unit tests passed (both Java and C-client). > > > - Checkstyle passed > > > - I executed a rolling-upgrade test from 3.5.7 to 3.6.1. (using > > > https://github.com/symat/zk-rolling-upgrade-test) > > > > > > Kind regards, > > > Mate > > > > > > On Thu, Apr 16, 2020 at 5:45 AM Patrick Hunt <[email protected]> wrote: > > > > > > > +1 - xsum/sig validated, rat ran clean, I was able to compile and ran > > some > > > > manual tests on varying cluster sizes. > > > > > > > > Patrick > > > > > > > > On Wed, Apr 15, 2020 at 11:44 AM Enrico Olivelli <[email protected]> > > > > wrote: > > > > > > > > > This is a release candidate for 3.6.1. > > > > > > > > > > It is a bugfix release and it introduces a few bugfixes and new > > features > > > > in > > > > > these areas: > > > > > - compatibility with applications built against 3.5 client libraries > > > > > (restored a few non public APIs) > > > > > - update Netty to 4.1.48.Final > > > > > - ability to pass configuration as file in zkCli for TLS config > > > > > - Add setKeepAlive support for NIOServerCnxn > > > > > - Fix server side request throttling > > > > > > > > > > The full release notes is available at: > > > > > > > > > > > > > > > > > > > > > https://issues.apache.org/jira/secure/ReleaseNote.jspa?projectId=12310801&version=12346764 > > > > > > > > > > *** Please download, test and vote by April 19th 2020, 23:59 UTC+0. > > *** > > > > > > > > > > Source files: > > > > > https://people.apache.org/~eolivelli/zookeeper-3.6.1-candidate-0/ > > > > > > > > > > Maven staging repo: > > > > > > > > > > > https://repository.apache.org/content/repositories/orgapachezookeeper-1056 > > > > > > > > > > The release candidate tag in git to be voted upon: release-3.6.1-0 > > > > > https://github.com/apache/zookeeper/tree/release-3.6.0-1 > > > > > > > > > > ZooKeeper's KEYS file containing PGP keys we use to sign the release: > > > > > https://www.apache.org/dist/zookeeper/KEYS > > > > > > > > > > The staging version of the website is: > > > > > > > > > > > https://people.apache.org/~eolivelli/zookeeper-3.6.1-candidate-0/website/ > > > > > > > > > > Should we release this candidate? > > > > > > > > > > Enrico Olivelli > > > > > > > > > > >
