Make the event number 64 bits and unsigned #398
Conversation
|
This breaks backwards compatibility for RNTuple as that does not have the necessary schema evolution support to do this on the fly (at least for RC2 of RNTuple). |
|
But I guess it's quite unlikely someone has RNTuples files and a file with event headers at the same time. So recreate the RNTuple file for version 0.99 with a 64 bit event number? It will be unreadable by EDM4hep 0.99, but is that even going to happen? |
Yes that is also what I would do. We should also go to the release RNTuple format there, i.e. ROOT >= 6.34, I think. |
Likely to run into issues with the range of 32 bit integers and the exepected event numbers. LHC is going beyond 32 bit range already in long runs.
jmcarcell
force-pushed
the
event-number-u64
branch
from
459ff94 to
8fd8cc6
Compare
There was a problem hiding this comment.
I think there is nothing left to do in this one. I think the check for the version of ROOT greater or equal than 6.34 is better than the previous check (below 6.33), which would have meant that the test would have stopped running forever until someone noticed. Related to this PR, maybe we could review some of the int32_t that are in other types in EDM4hep.
BEGINRELEASENOTES
EventHeader::eventNumbera 64 bit unsigned integer to avoid issues with the range of 32 bit integers.ENDRELEASENOTES
This is another small fix that came out of the discussion with HepMC folks (and their experience from LHC). We are likely to run into issues with the range of 32 bit integers and the exepected event numbers. LHC is going beyond 32 bit range already in long runs.