[3.8.x] [MNG-7476] Display a warning when an aggregator mojo locks other mojos executions

[gnodet]

This needs to be applied to 3.8.x, 3.9.x and master branches

[gnodet]

Fwiw, I've raised #737 to make the console output adapt to the terminal width.

[michael-o]

This now interferes/depends on the other PR, do you want to keep this one as-is or use the old style for now until the message formatting is properly defined/discussed?

[gnodet]

This now interferes/depends on the other PR, do you want to keep this one as-is or use the old style for now until the message formatting is properly defined/discussed?

I'd rather commit that one to 3.8.x and other branches asap and investigate the other PR for 3.9.x and master once this one is committed.

[michael-o]

This now interferes/depends on the other PR, do you want to keep this one as-is or use the old style for now until the message formatting is properly defined/discussed?

I'd rather commit that one to 3.8.x and other branches asap and investigate the other PR for 3.9.x and master once this one is committed.

Alright, will retest tomorrow.

[gnodet]

Fwiw, I've reviewed the original MNG-7433 issue and especially the reproducer and found out the issue was just a bug in the original project locking and not simply a missing warning. The lock was acquired too early, so that forked lifecycles could not obtain the lock for the project at all.

[michael-o]

@gnodet Can the reproducer be turned into an IT?

[gnodet]

@gnodet Can the reproducer be turned into an IT?

I suppose. The failure would lead to a deadlock. is there a timeout that can be set in the invoker ?

[michael-o]

@gnodet Can the reproducer be turned into an IT?

I suppose. The failure would lead to a deadlock. is there a timeout that can be set in the invoker ?

@slawekjaranowski Do you know this from top of your head?

[gnodet]

@gnodet Can the reproducer be turned into an IT?

@michael-o apache/maven-integration-testing#162

[slawekjaranowski]

@gnodet Can the reproducer be turned into an IT?

I suppose. The failure would lead to a deadlock. is there a timeout that can be set in the invoker ?

@slawekjaranowski Do you know this from top of your head?

There is no implementation for timeouts in Verifier.

[michael-o]

Moved message helper to #746.

[michael-o]

@gnodet Can the reproducer be turned into an IT?

I suppose. The failure would lead to a deadlock. is there a timeout that can be set in the invoker ?

@slawekjaranowski Do you know this from top of your head?

There is no implementation for timeouts in Verifier.

I have checked the source code of verifier. The lower level library supports timeouts in seconds, so we just need to wrap this with verifier. The JVM must be forked for this, so if timeout is > 0 and forkMode is auto, then autoforking must happen.

[michael-o]

Polished commits, they are now subsequently, standalone testable (tested actually). As with previous releases I have created a JIRA issue which addresses the problem which in turn solves the regression. Immediately visible in the changelog. This PR works decently with the fixed IT PR. Only one open question needs to be answered by @gnodet. Will align IT as well. Thank you for the effort.

So this does not require squashing, but addresses two subsequent issues. Just need upporting to 3.9.x and master.

[michael-o]

IT aligned and blessed.

[gnodet]

@michael-o I've slightly refactored the MojoExecutor to cleanup the access to the mojos map.

[michael-o]

@michael-o I've slightly refactored the MojoExecutor to cleanup the access to the mojos map.

This looks good and makes sense. Totally scoped in the ProjectLock. As far as I can see this commit shall be merged into "[MNG-7487]" and "[MNG-7476]" rebased on top of that.

[gnodet]

@michael-o I've slightly refactored the MojoExecutor to cleanup the access to the mojos map.

This looks good and makes sense. Totally scoped in the ProjectLock. As far as I can see this commit shall be merged into "[MNG-7487]" and "[MNG-7476]" rebased on top of that.

Done

[michael-o]

Fantastic, let's roll!