prefer use of InheritableThreadLocal to fix possible issue when mojo start a thread and try to access some artifacts

[olamy]

Signed-off-by: Olivier Lamy olamy@apache.org

Following this checklist to help us incorporate your
contribution quickly and easily:

• Make sure there is a JIRA issue filed
  for the change (usually before you start working on it). Trivial changes like typos do not
  require a JIRA issue. Your pull request should address just this issue, without
  pulling in other changes.
• Each commit in the pull request should have a meaningful subject line and body.
• Format the pull request title like [MNG-XXX] - Fixes bug in ApproximateQuantiles,
  where you replace MNG-XXX with the appropriate JIRA issue. Best practice
  is to use the JIRA issue title in the pull request title and in the first line of the
  commit message.
• Write a pull request description that is detailed enough to understand what the pull request does, how, and why.
• Run mvn clean verify to make sure basic checks pass. A more thorough check will
  be performed on your pull request automatically.
• You have run the Core IT successfully.

If your pull request is about ~20 lines of code you don't need to sign an
Individual Contributor License Agreement if you are unsure
please ask on the developers list.

To make clear that you license your contribution under
the Apache License Version 2.0, January 2004
you have to acknowledge this by using the following check-box.

• I hereby declare this contribution to be licenced under the Apache License Version 2.0, January 2004

• In any other case, please file an Apache Individual Contributor License Agreement.

[michael-o]

Merci! Do you think it is worth creating a unit test or IT to depict how it helps solving a problem?
I need to apply it to master first.

[michael-o]

Issue created: MNG-7212

[famod]

I suggest to merge #527 first (since it fixes a more severe problem) and iterate on this afterwards.
I can have a look at adjusting the test I introduced in #527 that will likely fail with InheritableThreadLocal.

[famod]

Btw, with this change, the orginal concurrency issue (that was addressed by #413) might come back in another form.
That is, if the spawned thread is modifying the artifact filter or the artifacts, another concurrently running thread might (again) end up with missing artifacts.
So I'm a bit on the fence about this. WDYT?

[famod]

Having said the above, as long as the original problem with aggregating goals is still fixed with change (I'll check later), I think it should be ok to merge.
If plugin or extension developers modify artifacts in a new sub thread, they have basically the same problem as before the introduction of the TL, so...

[michael-o]

@olamy Can this be closed in favor of #578 since the thread local solution will be reverted?

[olamy]

@michael-o yup sure do it. #578 just introduce a very high level (unmaintainable? really needed?) complexity.
There is still no real proof this simple change is bad.
but yeah (no time for argue) so it is.

[michael-o]

@michael-o yup sure do it. #578 just introduce a very high level (unmaintainable? really needed?) complexity. There is still no real proof this simple change is bad. but yeah (no time for argue) so it is.

To be honest, I cannot judge either. Can you add your opinion to #578, please? I'd really prefer to release 3.8.4 without and take more time to properly analyze the new proposal, but as you can see people only complain with releases.

[rmannibucau]

InheritableThreadLocal works only when subclassloader tree is controlled and can accept its value leakage. It is the case for jetty but not for all mojo so iy is not a good solution neither as a generic one. 578 has soem drawback but at least monothreaded case does not involve any of the regressions seen with last releases with is the minimum expectation IMHO.

[michael-o]

InheritableThreadLocal works only when subclassloader tree is controlled and can accept its value leakage. It is the case for jetty but not for all mojo so iy is not a good solution neither as a generic one. 578 has soem drawback but at least monothreaded case does not involve any of the regressions seen with last releases with is the minimum expectation IMHO.

Do you see #578 as mergeable at all or shall we seek out for a completely different solution?

[rmannibucau]

578 gets backs to an usable state for most case with a fully functional one with 1 thread so i would merge it yes.

[michael-o]

578 gets backs to an usable state for most case with a fully functional one with 1 thread so i would merge it yes.

And that superseded this PR for you?

[rmannibucau]

Yes

[famod]

I personally would prefer #476 over my PR #578 because it seems to be more efficient and catches more cases.
But I guess that #476 is a bit much of a global change for a 3.8.x? There could be a system property to opt out of it, just in case.
Anyway, if #578 is merged (for 3.8.4), #476 should be strongly considered for 3.9.0! (replacing #578)

[michael-o]

I personally would prefer #476 over my PR #578 because it seems to be more efficient and catches more cases. But I guess that #476 is a bit much of a global change for a 3.8.x? There could be a system property to opt out of it, just in case. Anyway, if #578 is merged (for 3.8.4), #476 should be strongly considered for 3.9.0! (replacing #578)

Ooph, I need to digest this first. Thanks for a proper analysis.

Maybe not to merge #578 at all in 3.8.x and not add something incomplete, but rather wait for 3.9.0 for #476?
Does #476 supersede #578?

[rmannibucau]

I would agree with #476 but didn't get much time to test but if you are confident it is better than any thread local solution anyway so happy to go with it too.

[michael-o]

As far as I understood this discussion this is superseded by the two other PRs.