Hi, this error helps. It looks like the lucene-core.jar file does not have the META-INF/services. I wonder how Solr works!
It would be good to figure out what's going on and why the JAR files in your local repository do not have the required META-INF/services subfolder. Uwe ----- Uwe Schindler Achterdiek 19, D-28357 Bremen https://www.thetaphi.de eMail: [email protected] > -----Original Message----- > From: Michael Gibney <[email protected]> > Sent: Wednesday, June 2, 2021 6:05 AM > To: [email protected] > Subject: Re: Solr/Lucene joint development workflow? > > Uwe, > Thanks for this advice! I followed this as closely as I could > (including the gradle `--refresh-dependencies` flag), and was again > able to compile (and compileTest) against mavenLocal. But on actually > running tests, afaict all tests fail with the same stacktrace (copied > below). iirc this was the same error I was getting previously with the > SNAPSHOT build (before the introduction of the prerelease dependency). > I'd be grateful for any advice on addressing this remaining error (and > having encountered the error, I continue to be curious whether anyone > has this workflow or something similar working in practice). > > David, your suggestion to add "mavenLocalI()" to Gradle's repo list > seems reasonable -- perhaps a more viable option now that the main > branch is pegged to specific builds (as opposed to a SNAPSHOT)? I'm > happy to help contribute to documentation of this process (once I have > it working!). > > One other challenge I encountered was determining which commit the > current prerelease build is based on. The desired workflow would be to > update the Lucene feature branch by merging upstream changes as far as > the commit of the prerelease build (for use as a dependency of the > Solr-side feature branch). > > The exception I'm getting on running tests is: > > java.lang.ExceptionInInitializerError > at org.apache.lucene.codecs.Codec.getDefault(Codec.java:141) > at > org.apache.lucene.util.TestRuleSetupAndRestoreClassEnv.before(TestRuleSetup > AndRestoreClassEnv.java:133) > at > org.apache.lucene.util.AbstractBeforeAfterRule$1.evaluate(AbstractBeforeAfter > Rule.java:42) > at > com.carrotsearch.randomizedtesting.rules.StatementAdapter.evaluate(Stateme > ntAdapter.java:36) > at > org.apache.lucene.util.TestRuleStoreClassName$1.evaluate(TestRuleStoreClass > Name.java:38) > at > com.carrotsearch.randomizedtesting.rules.NoShadowingOrOverridesOnMethod > sRule$1.evaluate(NoShadowingOrOverridesOnMethodsRule.java:40) > at > com.carrotsearch.randomizedtesting.rules.NoShadowingOrOverridesOnMethod > sRule$1.evaluate(NoShadowingOrOverridesOnMethodsRule.java:40) > at > com.carrotsearch.randomizedtesting.rules.StatementAdapter.evaluate(Stateme > ntAdapter.java:36) > at > com.carrotsearch.randomizedtesting.rules.StatementAdapter.evaluate(Stateme > ntAdapter.java:36) > at > org.apache.lucene.util.TestRuleAssertionsRequired$1.evaluate(TestRuleAssertio > nsRequired.java:53) > at > org.apache.lucene.util.AbstractBeforeAfterRule$1.evaluate(AbstractBeforeAfter > Rule.java:43) > at > org.apache.lucene.util.TestRuleMarkFailure$1.evaluate(TestRuleMarkFailure.ja > va:44) > at > org.apache.lucene.util.TestRuleIgnoreAfterMaxFailures$1.evaluate(TestRuleIgn > oreAfterMaxFailures.java:60) > at > org.apache.lucene.util.TestRuleIgnoreTestSuites$1.evaluate(TestRuleIgnoreTest > Suites.java:47) > at org.junit.rules.RunRules.evaluate(RunRules.java:20) > at > com.carrotsearch.randomizedtesting.rules.StatementAdapter.evaluate(Stateme > ntAdapter.java:36) > at > com.carrotsearch.randomizedtesting.ThreadLeakControl$StatementRunner.run( > ThreadLeakControl.java:370) > at > com.carrotsearch.randomizedtesting.ThreadLeakControl.lambda$forkTimeoutin > gTask$0(ThreadLeakControl.java:826) > at java.base/java.lang.Thread.run(Thread.java:834) > > Caused by: > java.lang.IllegalArgumentException: An SPI class of type > org.apache.lucene.codecs.Codec with name 'Lucene90' does not exist. > You need to add the corresponding JAR file supporting this SPI to your > classpath. The current classpath supports the following names: [] > at > org.apache.lucene.util.NamedSPILoader.lookup(NamedSPILoader.java:113) > at org.apache.lucene.codecs.Codec$Holder.<clinit>(Codec.java:58) > ... 19 more > > > On Fri, May 28, 2021 at 4:03 PM David Smiley <[email protected]> wrote: > > > > To whomever does this first, it would be great if the steps could be written > down carefully in more detail so that it can be shared for its eventual > inclusion > in /dev-docs > > > > Maybe "mavenLocalI()" should simply be added to Gradle's repo list by > default? This will speed up some first-time builds in some environments who > already have lots of local Maven dependencies. > > > > ~ David Smiley > > Apache Lucene/Solr Search Developer > > http://www.linkedin.com/in/davidwsmiley > > > > > > On Thu, May 27, 2021 at 12:09 PM Uwe Schindler <[email protected]> > wrote: > >> > >> Hi Michael, > >> > >> since yesterday, the Lucene dependency is no longer a snapshot one, so > parts of your mail are no longer valid. But the worksflow is very similar. > >> > >> If the Solr team wants to use a newer preview build of Lucene, there is a > workflow using ASF Jenkins that can build a new "prerelease" release of the > Lucene main branch and deploy it on some ASF data dump server as maven > repository. > >> > >> The Gradle build of Solr is referring to the recent repository and picks a > specific version (the build number of ASF jenkins). > >> > >> For joint development, I'd suggest to work like this: > >> - Change the dependency in the gradle versions.props file to SNAPSHOT > (only local, don't submit this as PR) > >> - Add mavenLocal() as suggested in your mail (only local, don't submit > >> this as > PR) > >> - develop changes in lucene and install them in local repo > >> - use them from solr through the snapshot dependency. BUT: make sure the > snapshot is updated, this can be enforced by passing a gradlew command line > parameter to redownload all dependencies. > >> - once you are done with joint development create pull requests in both > repositories > >> - Lucene's change should be merged first. Once this is done and all tests > pass, ask a committer to trigger the Jenkins job to build a new prerelease > >> - Update the Solr pull request and enter the new repository coordinates > created by Jenkins (needs to be done in 2 files, see below). > >> > >> For more details how to update Lucene dependencies in Solr, run: > >> $ gradlew :helpDeps > >> > >> Uwe > >> > >> ----- > >> Uwe Schindler > >> Achterdiek 19, D-28357 Bremen > >> https://www.thetaphi.de > >> eMail: [email protected] > >> > >> > -----Original Message----- > >> > From: Michael Gibney <[email protected]> > >> > Sent: Wednesday, May 26, 2021 8:02 PM > >> > To: [email protected] > >> > Subject: Solr/Lucene joint development workflow? > >> > > >> > I'm working on some features that involve changes to both Lucene and > >> > Solr. Post-TLP-split, I'm wondering whether anyone has recommended > >> > techniques to handle this kind of situation. > >> > > >> > Ideally one would work on Lucene changes, get them merged, and then > >> > proceed with Solr development; but realistically even if this were as > >> > straightforward in practice as it sounds in principal, there are cases > >> > where one would still want to develop in parallel. > >> > > >> > I haven't been able to find any documented recommendation on this > >> > subject. It's possible to have a locally built Lucene snapshot (via > >> > `gradlew mavenToLocalRepo`); but I was only able to actually _build_ > >> > Solr against the local Lucene artifact by adding `mavenLocal()` to the > >> > `allprojects/repositories` block in `gradle/defaults.gradle` -- and I > >> > have yet to figure a way get the local Lucene artifact on the test > >> > classpath (so I'm as yet unable to run Solr tests that depend on > >> > unmerged upstream changes to Lucene). > >> > > >> > It's also possible that the partially-functional approach described > >> > above will have to change now that Solr main depends on a specific > >> > Lucene snapshot version. > >> > > >> > Is anybody doing something like this? Or perhaps I'm asking the wrong > >> > question? I can think of solutions that involve setting up my own > >> > maven repository, to which I publish my own pinned versions of Lucene, > >> > and refer to such pinned versions/repo as part of a given Solr > >> > "patch". But that feels both half-baked _and_ bloated, so I don't want > >> > to go down that road unless I feel convinced there's no better > >> > alternative. > >> > > >> > Michael > >> > > >> > --------------------------------------------------------------------- > >> > To unsubscribe, e-mail: [email protected] > >> > For additional commands, e-mail: [email protected] > >> > >> > >> --------------------------------------------------------------------- > >> To unsubscribe, e-mail: [email protected] > >> For additional commands, e-mail: [email protected] > >> > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
