[ntuple] `RNTupleFillContext`: Wait for all tasks before destructing model

[hahnjo]

Re-enable the SmallClusters test.

[phsft-bot]

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default How to customize builds

[phsft-bot]

Build failed on mac12arm/cxx20. Running on 194.12.161.128:/Users/sftnight/build/workspace/root-pullrequests-build See console output.

Failing tests:

• projectroot.tree.ntuple.v7.test.gtest_tree_ntuple_v7_test_ntuple_extended

[github-actions[bot]]

Test Results

    13 files      13 suites   2d 18h 38m 51s :stopwatch:  3 028 tests  3 028 :white_check_mark: 0 :zzz: 0 :x: 33 838 runs  33 838 :white_check_mark: 0 :zzz: 0 :x:

Results for commit 74ea29c1.

:recycle: This comment has been updated with latest results.

[phsft-bot]

Starting build on ROOT-performance-centos8-multicore/soversion, ROOT-ubuntu2204/nortcxxmod, ROOT-ubuntu2004/python3, mac12arm/cxx20, windows10/default How to customize builds

[hahnjo]

Awesome 😃! Does this finally fix the crashes in some build configurations?

Yes. But we're not entirely happy with the solution because it breaks some of the layering (IMHO RNTupleWriter should not need to know that the sink has to wait for all its tasks). We're still discussing how to best solve this...

[hahnjo]

https://github.com/root-project/root/pull/16164 nicely solved the problem with freeing pages, but the RNTupleModel problem persists because the tasks created in RPageSinkBuf hold onto the RColumnElement...

[hahnjo]

This isn't needed anymore, https://github.com/root-project/root/pull/16350 removed the small clusters write option.