[
https://issues.apache.org/jira/browse/MBUILDCACHE-105?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17876885#comment-17876885
]
ASF GitHub Bot commented on MBUILDCACHE-105:
--------------------------------------------
AlexanderAshitkin commented on code in PR #176:
URL:
https://github.com/apache/maven-build-cache-extension/pull/176#discussion_r1732039886
##########
src/main/java/org/apache/maven/buildcache/CacheControllerImpl.java:
##########
@@ -255,6 +255,11 @@ private CacheResult analyzeResult(CacheContext context,
List<MojoExecution> mojo
build.getCacheImplementationVersion());
}
+ if (lifecyclePhasesHelper.isLaterPhaseThanBuild("package", build))
{
+ LOGGER.warn("Cached build doesn't include phase 'package',
cannot restore");
Review Comment:
The expected behavior is that when two consecutive 'mvn compile' commands
are used, the second 'mvn compile' should use the cache. I'm concerned that
this implementation might lead to a performance regression. A more accurate
implementation would be to always attach the generated sources in `save` and
process partial success more accurately.
In the current implementation, the 'analyzeResult' returns partial success
here (see lines 277-278) as expected. The fix should be made in the partial
success processing logic. A build without a package phase should restore
generated sources only and rerun all the missing lifecycle phases to build
artifacts.
> Executing 'clean process-test-classes' then 'clean verify' leads to empty
> jars
> -------------------------------------------------------------------------------
>
> Key: MBUILDCACHE-105
> URL: https://issues.apache.org/jira/browse/MBUILDCACHE-105
> Project: Maven Build Cache Extension
> Issue Type: Bug
> Reporter: Réda Housni Alaoui
> Priority: Major
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
