Custom member streamer for CalArray<o2::tpc::PadFlags>::mData

[sawenzel]

This commit expands on #14427 and fixes the issue brought up in https://its.cern.ch/jira/browse/O2-4671.

After debugging/testing it turns out that the approach taken via a customer streamer for std::vector<o2::tpc::PadFlags> does not take effect in the ROOT/IO because apparently ROOT still prefers to use the CollectionProxy for std::vector and does not employ the custom streamer.

Instead, after discussion with @pcanal, this commit proposes to implement a custom stream just for the mData data member of CalArray<o2::tpc::PadFlags>. This is the only place where we use o2::tpc::PadFlags in IO and it fixes the problem when reading CCDB objects containing such data.

I have verified that the following code

o2-ccdb-downloadccdbfile -p TPC/Calib/IDC_PadStatusMap_A -t 1731274461770 -d ./ -o tpc_idc.root --no-preserve-path
root tpc_idc.root
gFile->Get<o2::tpc::CalDet<o2::tpc::PadFlags>>("ccdb_object")

correctly executes the custom streamer function.

Note that there is also no need to make the code ROOT version dependent. We need to fix the reading in any case and the writing will just stay the same.

Concerning situations, where future classes will write data containing std::vector<o2::tpc::PadFlags> we should be protected by the fact that this bug has been fixed >= ROOT 6.36 in any case.

This commit relates also to

root-project/root#17009

[github-actions]

REQUEST FOR PRODUCTION RELEASES:
To request your PR to be included in production software, please add the corresponding labels called "async-" to your PR. Add the labels directly (if you have the permissions) or add a comment of the form (note that labels are separated by a ",")

+async-label <label1>, <label2>, !<label3> ...

This will add <label1> and <label2> and removes <label3>.

The following labels are available
async-2023-pbpb-apass4
async-2023-pp-apass4
async-2024-pp-apass1
async-2022-pp-apass7
async-2024-pp-cpass0
async-2024-PbPb-apass1
async-2024-ppRef-apass1
async-2024-PbPb-apass2
async-2023-PbPb-apass5

[ktf]

As discussed privately, I think indeed the current code right now works by accident in certain builds due to the "static initialisation order fiasco" accidentally being the one we need. Given this improves over it, I am all for it.

[sawenzel]

The problem is not related to static initialization but rather that ROOT ignores custom streamer functions on std C++ collections. The following code demonstrated this:

root [0] auto f=[](TBuffer& R__b, void* objp) { std::cout << "Hello\n"; return; };
root [1] std::vector<int> vec(10);
root [2] auto cl = TClass::GetClass<std::vector<int>>();
root [3] cl->SetStreamerFunc(f);
root [4] TFile out("outfile.root", "CREATE");
root [5] TTree tree("data","data");
root [6] tree.Branch("branch", &vec);
root [7] tree.Fill();

because the writing of vec does not print anything. However, using custom streamers on the level of members of C++ classes works fine, which is done with this PR.

[ktf]

I am puzzled. I developed the code with a similar small case reproducer and there is a different behavior in O2Physics with my previous patch removed. Do you see the same without the pragma link C++ class std::vector < o2::tpc::PadFlags> + ?

[sawenzel]

Yes I tried with Philippe the cases with and without the #pragma and it didn't make a difference.