Clean up Trigger ByteStream reading configuration
The main goal of this MR is to get rid of a copy-paste with very long history that is the content of the file jobOfragment_ReadBS_standalone.py
. None of the options set there are needed / valid / justified right now. If anything from these lines would be needed, the clients needing it should take care of the correct configuration. Several related changes are made to clean up the configuration and make everything work.
The specific changes and their relations are:
- The file
jobOfragment_ReadBS_standalone.py
is removed. - The exact copy of these lines inside
ByteStreamConfig.py
is also removed, making the functionTrigBSReadCfg
redundant. Since it is used by many clients, it is left as an alias ofByteStreamReadCfg
for now. Clients will be adapted in a separate MR and theTrigBSReadCfg
will be removed then. - Remove EventInfo from InitCnvs in
ByteStreamReadCfg
- the legacy EventInfo BS converter was deleted months ago and this line must've been missed and was causing unnecessary warnings in every job reading BS. - Call to old-style BS reading config in
runHLT_standalone.py
is replaced with new-JO fragment (ByteStreamReadCfg
fromByteStreamConfig.py
) to keep the Run-3 configuration consistent and up to date. This involved a small move of flags import inrunHLT_standalone.py
, so cc @stelzer. - To make 4. work, I had to extend the functionality of the
CAtoGlobalWrapper
to also merge AppMgr properties. This also fixes ATR-21307. cc @tbold, @wlampl - Follow-up from 1: There was only one client relying on one line from
jobOfragment_ReadBS_standalone.py
, namelyTrigMuonRoITool
. The way it works relies on non-MT-friendly solution involving Gaudi Converter for MUCTPI_RDO which was hidden in the configuration. I wrote a new configuration fragment for the tool where this dependency is explicit and comments are left explaining what should be done in the future for MT-friendly solution. cc @sshaw, @wiedenma - Follow-up from 2: Move some clients in TileCalorimeter to use
ByteStreamReadCfg
with explicitly naming the required types instead of using the oldTrigBSReadCfg
Flagging this as full-unit-tests because of the changes in central parts of configuration.
Edited by Rafal Bielski