- Move all configuration of athenaHLT services to TriggerUnixStandardSetup.py.
  Remove the use of includes as much as possible (apart from user job options).
- Replace the OfflineTHistSvc.py include with a simple flag
- Add error in case the offline THistSvc is used with more than one child
- Move the configuration of HLTResultMTMaker to its own module

Files with identical content (same SHA1) but different file permissions
cause confusing differences shown in GitLab. The file permissions of
61 files were set to 644 (ATLINFR-2011).

Charles Leggett authored 6 years ago and Vakhtang Tsulaia committed 6 years ago

Former-commit-id: 645176ec

Add `--concurrent-events` command line arg to set the number of slots.
Also make sure the relevant jobproperties are set on the athena side.

Also switch to using the thread-safe `DFDcmEmuBackend`.


Former-commit-id: 05df8bde

`TrigISHelper` was mainly used to read the magnet currents from IS.
These are since quite some time provided via the configuration ptree and
hence this helper tool has no use anymore. In general, we shouldn't be
reading from IS in the HLT anyway.


Former-commit-id: 717ce229

- `HltOstream`: not used anywhere
- `[I]TrigIS`: never used and we have a better way to publish to IS
- `IHltTHistSvc`: we never used this HLT-specific interface


Former-commit-id: a2131908

For a simple HelloWorld example use
`AthExHelloWorld/HelloWorldOptions.py` instead.


Former-commit-id: f7d57492

This package is obsolete and hasn't been used since Run-1. For similar
features, see the TrigExPartialEB package.


Former-commit-id: e75de763

Frank Winklmeier authored 6 years ago and Walter Lampl committed 6 years ago

Former-commit-id: 03fa8b08

Added try-catch blocks around hltinterface and eformat calls where missing and improved
the existing handling. Removed include of EventId, HLTResult and DataCollector headers
where not used anymore.


Former-commit-id: 197f35fc

The PSC modifies its state machine if we are running in interactive mode
(to allow debugging of the python configuration before initialize is
called). Previously interactive mode was flagged by a global
'interactive' variable. This is now done directly in athenaHLT.py.


Former-commit-id: b7139344

- Remove unnecessary `release()` calls (the ServiceHandle does this)
- Remove commented code
- Remove code to handle user commands as this is non-functional now
- Remove dllName, factoryName and EvtSel properties (see HLTPUDal!1)


Former-commit-id: 355f8d67

Former-commit-id: 8dfa8e1c

Adapt asetup_wrapper to latest version of asetup. Add test and fix
handling of preloading and extra_sw_path.


Former-commit-id: 98e2d150

Since the migration of the asetup_wrapper, we no longer need the scripts
to create a static HLT-Environment.data.xml file.


Former-commit-id: 8be669d8

Former-commit-id: f89609cf

Test was timing out in some nightlies.



Former-commit-id: a97b4bee

Changes include:
* implement HLTResultMTByteStreamDecoderAlg and the corresponding tool -> this reentrant algorithm replaces the functionality of Gaudi Converter for BS decoding, since Converters are inherently not compatible with athenaMT
* make the Converter method HLT::HLTResultMTByteStreamCnv::createObj fail when used with a message to use the Alg instead
* allow TriggerEDMDeserialiserAlg to skip events with no payload - this is needed at least for testing our first bytestream files with new HLTResultMT
* rename one of the HLTResultMT::setStatus overload to setErrorCodes because the overload was ambiguous (compilation error when used)
* add the implementation of operator<<(std::ostream&, const HLT::HLTResultMT&) for useful debug printouts
* add MTCalibPebDecisions to the list of serialised collections in MTCalibPeb.py in order to produce some payload
* fix the payload writing in HLT::HLTResultMTByteStreamCnv::createRep (missing reference operator)


Former-commit-id: e03390a9

Former-commit-id: c7f94bcb

- Add missing IncludeError import
- Migrate ScopeTimer to std::chrono
- General cleanup


Former-commit-id: 96e07a3e

Unit tests are not working yet until we have an updated version of
HLTMPPU in the tdaq release.


Former-commit-id: 70521c4c

Former-commit-id: 0b2826eeaf38b21951809a297d4a594c7ee6202e

also allow configuring ROBs and SubDets in stream tags made by StreamTagMakerTool


Former-commit-id: 9d45aefa

Former-commit-id: f7a0e950

- Remove hard-coded OutputLevels
- Do not rely on global OutputLevel/HLTOutputLevel variables


Former-commit-id: 6fd20258

- The python OutputLevel is now set in athenaHLT.py
- The showInclude and tracePattern OKS attributes have been removed
- Remove the global `OutputLevel` and `HLTOutputLevel` variables


Former-commit-id: 545e368a

The previous athenaHLT.py from HLTTestApps is now replaced by a simple
wrapper script around runHLTMPPy. Most options are the same as before,
but some were removed, some renamed for consistency.


Former-commit-id: e011aab9

Former-commit-id: b087b026

also change the tool input from a vector to a single object (no need for multiple inputs)


Former-commit-id: 6ec3aabd

Deprecating use of non-bound DataHandle, especially non-const.


Former-commit-id: fea4eb13

* Add some error handling in HLTResultMT introducing StatusCode and AthCheckMacros
* HLTResultMT now removes duplicate stream tags also in setStreamTags - both setStreamTags and addStreamTag now return StatusCode
* HLTResultMT::getSerialisedData now returns StatusCode and pointer to the data is passed in the argument - this ensures error handling in the case when the caller asks for data with a moduleId which is not in the stored map
* new method HLTResultMT::addSerialisedDataWithCheck to add data only for modules which are not yet in the map - returns StatusCode::FAILURE if moduleId is already in the map; the no-check method void addSerialisedData simply appends new data if moduleId already exists
* Ensure the status words in HLTResultMT consist of 1 bit-mask word (eformat::helper::Status) + optional error codes (used to store hltonl::PSCErrorCode). This is how the status field was used in Run 2, although the first word was built entirely on the TDAQ side in DCM. The agreement between HLT and TDAQ is to provide it in this form already from the HLT side in the new interface where we produce a FullEventFragment, then DCM can simply apply an OR bit mask to the first word without touching the rest, if it wants to add extra error flags.
* HLT bits are now stored as boost::dynamic_bitset, which makes the bit operations less error prone, e.g. HLTResultMT::addHltBit is now well defined and easy to follow. For serialisation purpose, HLTResultMT still needs to own a vector<uint32_t> representation of the bits because the lifetime of this vector has to extend beyond the BS converter createRep method, namely until the end of conversion service's commitOutput. The vector representation does not have any setters and is updated only by the getter method.
* Add more comments in HLTResultMT.h
* Adapt TriggerBitsMakerTool, TriggerEDMSerialiserTool, TriggerEDMDeserialiserAlg, HLTResultMTByteStreamCnv, HltEventLoopMgr to the changes
* Improve formatting of FullEventFragment header printout in TrigByteStreamCnvSvc


Former-commit-id: 32678df6

Rafal Bielski authored 6 years ago and Walter Lampl committed 6 years ago

Former-commit-id: 92b5986d

Update packages:HLToks JiveXMLServer HLTUtils TrigTransform TrigCommon TrigExMTHelloWorldLvl1 TrigExPartialEB TrigExL2muCalibTest TrigExMTHelloWorld TrigServices TrigPython TrigKernel TrigPSC HLTRates2COOL TrigOnlineMonitor TrigConf2COOLOnline HLTTestApps ByteStreamEmonSvc TrigByteStreamCnvSvc  from upstream/21.1 to upstream/master via pseudo-merge


Former-commit-id: 0fe4136b

also moving it back to the HLT namespace


Former-commit-id: 9c14c246

Compare type and name, and if the same tag exists, merge robs and dets.
Remove extra duplicate handling from callees.


Former-commit-id: 097df635

Move the functionality from HLTResultMTMakerAlg to HLTResultMTMaker and make the alg only call the tool.
Adjust the job options accordingly.


Former-commit-id: 69f08f09

Former-commit-id: a938135f

Events with PSCErrorCode::PROCESSING_FAILURE, i.e. those with algorithm errors
reported by AlgExecStateSvc, should go into a different debug stream from
events with framework errors, i.e. failures in HltEventLoopMgr


Former-commit-id: d69ae7d5