On 23/08/2013, at 2:47 AM, johnrengelman <john.r.engel...@gmail.com> wrote:
> I've started looking at a fix for GRADLE-2034 > <http://issues.gradle.org/browse/GRADLE-2034> with respect to mavenLocal() > having artifact POMs but not the corresponding artifact. It appears the > Maven does this a lot. I've been seeing it anytime I execute a 'mvn install' > on the same machine that I run Gradle builds on (totally prevents our CI > server being used for both Maven and Gradle). > > This is my plan to fix: > 1) Create a new 'DefaultMavenLocalArtifactRespository' extending from > 'DefaultMavenArtifactRespository' > 2) Override the 'createRealResolver' method to return a new > 'MavenLocalResolver' class > 3) 'MavenLocalResolver' extends 'MavenResolver' and overrides the > 'getDependency' method. > 4) 'MavenLocalResolve.getDependency' delegates to 'MavenResolver' for > snapshot behavior, otherwise.. > 5) Implement a new method 'resolveIfArtifactPresent' that upon 'ivyRef != > null', it calls the 'findAnyArtifact' method in 'ExternalResourceResolver' > and only resolves the artifact if the return is non-null > > I've implemented this fix locally and it appears to the solve the issue and > allow the mavenLocal resolve to fail-over to the next repository when it > finds a POM but no artifact. I'm implement some tests around these scenarios > now...I suspect a bunch of integration tests will break because they create > a Pom file test test resolution but not an artifact. > > Does this sound like a good plan for fixing the bug? Thanks for tackling this bug. Your plan sounds good. I would probably get rid of step 4, as I don't think there's any difference between a snapshot and a release as far as this bug goes - it affects anything that may have come from a remote repository. I'd also probably do step 5 only if the packaging is not 'pom'. If we find a pom whose packaging is 'pom', then we should probably use it. One question is what happens if I manually configure a `maven()` repository to point at the m2 cache? Or, if I manually configure a `mavenLocal()` so that it no longer points at the m2 cache? Possibly we want to decide whether to use the 'cache' (the new) or 'repository' (the current) behaviour based on whether a repository points at the m2 cache, regardless of how it ended up pointing there. -- Adam Murdoch Gradle Co-founder http://www.gradle.org VP of Engineering, Gradleware Inc. - Gradle Training, Support, Consulting http://www.gradleware.com Join us at the Gradle eXchange 2013, Oct 28th in London, UK: http://skillsmatter.com/event/java-jee/gradle-exchange-2013
