WIP Added schema evolution in BS

This MR adds schema evolution for BS serialsiation. Also syntetic tests are added. Tagging @ssnyder

marked as a Work In Progress

changed title from Added test for schema evolution in BS to WIP Added schema evolution in BS

changed the description

added 210 commits

• 0dee6bdf...02f41443 - 209 commits from branch atlas:master
• 2bd71cd7 - reduced scope of testing in bs reading test

Compare with previous version

added 1 commit

• 8c61411e - fixed tests precedence

Compare with previous version

added 1 commit

• 091f9820 - enable actual test

Compare with previous version

Hi @tbold,

just a friendly ping so this MR is not forgotten.

Cheers, Tomas (L1)

I did not. Just waiting (very) patiently for feedback from @ssnyder

Didn't realize you were waiting... you should go ahead with this.

Hi @ssnyder, Sorry to bug you. I can go ahead but only with "commented out" code actually triggering the schema evolution. The thing is that the code I wrote does not do the job evn if it is a mirror of existing code. I am not sure if I understand how the old code actually worked. Cheers, Tomasz

Ah --- you're not actually converting the data anywhere!

The way i did it relied on the implicit conversions done by root, by passing the desired type into ReadObjectAny via TrigTSerializer. That's why the aux type handling happens before the derserialization. In your code, it's the other way round, which makes it much more awkward.

Thanks, That makes it clear. So far we wrote 0 bytes of data with this so it can all be changed. I need now to check with @gemmeren if the "offline" serializer is equally capable. The exact code doing deserialisation is this one: https://gitlab.cern.ch/atlas/athena/blob/master/Database/AthenaPOOL/AthenaPoolCnvSvc/src/AthenaRootSerializeSvc.cxx#L83

Would ReaObjectAny be needed there?

Hi @tbold, @ssnyder, I guess there is still more discussion needed before this MR can go ahead? Thanks, Catrin (Thu morning's L1 shifter)

closed