[MNG-6114] Elements from the global settings should be ordered before elements from the user settings

[jira-importer]

Christian Schulte opened MNG-6114 and commented

From a logical point of view, in a stacked design we first have elements from global settings, physically followed user settings if they don't override global settings. Now, it is the opposite which is not logical.

---

Affects: 3.6.3

Issue Links:

• MNG-7018 Maven does not traverse all repos when downloading extensions from .mvn/extensions.xml
  ("causes")

Remote Links:

• Page
  
• GitHub Pull Request #142
  
• GitHub Pull Request #92
  
• GitHub Pull Request #142
  
• GitHub Pull Request #418
  

Backported to: 4.0.0-alpha-2

0 votes, 9 watchers

[jira-importer]

Christian Schulte commented

See this thread on dev@.

[jira-importer]

Stephen Connolly commented

Maven 3.4.0 has been dropped. See this thread for more details.

This issue will need to be re-scheduled for a Maven release in the (hopefully near) future.

[jira-importer]

Christian Schulte commented

Issue is fixed in the master branch of my private maven repository on github. When interested, you can cherry pick the commit to your own repository and create a pull request for the apache master branch with just this commit yourself from there.

[jira-importer]

Michael Osipov commented

Christian Schulte, do you have an IT at hand? Your fork does not contain any.

[jira-importer]

Michael Osipov commented

Robert Scholte, is this something we want in 3.7.0? It seems to be the right choice for me.

[jira-importer]

Michael Osipov commented

I have ran Core ITs with this enabled and it triggers a bug in Maven. Will create shortly.

[jira-importer]

Michael Osipov commented

Robert Scholte, Karl Heinz Marbaise, WDYT? This is one ugly.

[jira-importer]

Robert Scholte commented

You're saying that right now global settings is winning from user settings when using the same id?

[jira-importer]

Michael Osipov commented

Not directly, it is winning simply because it is the first on the list now which makes logically sense. See also MNG-5984.

[jira-importer]

Robert Scholte commented

Please update the description, I'm missing the bigger picture.

[jira-importer]

Robert Scholte commented

Just to confirm I understand: global has repos A and B, user has B and C

Current order is A (global) B (user) C(user)? Whereas you expect B(user) C(user) A(global)?

[jira-importer]

Michael Osipov commented

The exact opposite is the case: Current order is B (user) C (user) A (global). Whereas you expect A (global) B (user) C (user). See also my explanation in MNG-7018 as well as the settings files there.

[jira-importer]

Michael Osipov commented

But this change applies to all elements in the settings file if you look at the changeset closely.

[jira-importer]

Robert Scholte commented

A (global) B (user) C(user) was indeed my first expectation: first load global, user should override and expand. If that's not the case it makes sense to fix this in 4.0.0

[jira-importer]

Michael Osipov commented

OK, you have the same expectation. This is currently not met by the settings merger. I am currently thinking how MNG-7018 can be fixed. Christian Schulte worked out a patch which needs to be cherry-picked and properly merged.