Administrator authored 2 years ago and Pierre-Alain Loizeau committed 2 years ago

Restructure code in the main CMakeLists.txt

Administrator authored 2 years ago and Pierre-Alain Loizeau committed 2 years ago

Move code which does some basic checks into separate CMake file.
Some additional reordering of the code is needed.

Administrator authored 2 years ago and Pierre-Alain Loizeau committed 2 years ago

Move code which defines the compiler settings into separate CMake file.

Administrator authored 2 years ago and Pierre-Alain Loizeau committed 2 years ago

Move code which defines additional targets into separate CMake file.
Move code which generates the config files into separate CMake file.

Administrator authored 2 years ago and Pierre-Alain Loizeau committed 2 years ago

Check if environmnet variables SIMPATH and FAIRROOTPATH are set or if the paths
are passed on the commandline as -DSIMPATH=<path to FairSoft installation>
-DFAIRROOTPATH=<path to FairRoot installation> early during the cmake run.
After the initial check only the CMake variables are used.

Administrator authored 2 years ago and Pierre-Alain Loizeau committed 2 years ago

Don't require environmnet variables SIMPATH and FAIRROOTPATH when building
flesnet and NicaFemto. Pass the corresponding CMake variables to the build
processes of the packages.

Administrator authored 2 years ago and Pierre-Alain Loizeau committed 2 years ago

Remove obsolete, unneeded and unused code.

Dominik Smith authored 2 years ago and Pierre-Alain Loizeau committed 2 years ago

Unpack Much: Applied Epoch cycle fix. Fixed MacOS compilation (variable-sized array). Fixed missing Elink parameters.

- Prevent FairrootManager from touching the source object trhogh its pointer after the run is over
- Explicitely delete the dynamically allocated Run and Source objects in the macro (give up share ownership of config objects)
- Give up ownership of the config objects, which triggers their destruction before reaching end of the macro (last owner of shared pointer)

Problem was that the FairrootManager was never going out of scope and was not calling the Source destructor.
So all shared pointers and the source objects were destroyed only when going out of scope in the ROOT session, at which time ROOT already deleted some TF1's, THx's and Canvases without asking anybody (global Lists).
The source and config destructors were then calling destructors for their members and member's members, where we tried to cleanup the memory we assigned but which ROOT already deleted.

    - Change HWM for Sink process input to 1 as this applies to each link to an Evt builder individually and not to the sum of all as expected (solves huge bufferings)
    - Do not re-order the TS in the Sink by default (leads to buffering if one branch is really slow)
    - Add option for DigiEvent I/O between builders and Sink
    - Reduce HWM at input and output of Unpackes and Event builders to 1
    - Add bash and sbatch scripts for testing replay of a run on the mFLES cluster (single source node + single processing node)
    - Add information about online MQ problems in Gold run and these fixes

=> needed to collect histo and canvases from multiple instances of the devices (e.g. multiple nodes) in a single histo server
=> Same cannot easily be in processing classes like Unpack and BuildEvent as both name and folder need to be changed to avoid collisions

    - Fix to the performance plots of the DigiEventSink by changing the logic to Profile or new counts addition
    - Add option for DigiEvent I/O between builders and Sink

- Evolution of the full TS buffer size as function of device runtime
- Evolution of the missed TS buffer size as function of device runtime
- Evolution of the Processed full TS count as function of device runtime
- Evolution of the Processed missing TS count as function of device runtime
- Evolution of the Total processed TS count as function of device runtime
- Evolution of the Processed events count as function of device runtime

[MQ] in DigiEventSink, add flag to control Missing TS empty writing if consecutive check is ON, false by default

- The check for consecutive TS indices was still called when receiving Missing TS Data messages even if the bypass flag was set
  => Surround the call with proper if condition
- The range when taking the digis data out of the unpacked timeslice to the DigiEvent was missing the last digi for all detectors
  => Increment the stop iterator to make sure it is included
- In rare case, the device can be pulled down before the Init method is called/finished, which led to a segfault
  => add an InitDone flag + use it to control delete operations in the destructor

- Bring in new mapping setter from 819c026f
- More Default time offsets values based on RunId, Nickel ones are relative to BMon instead of TOF