Simone Sottocornola authored 3 weeks ago and Bertrand Martin Dit Latour committed 3 weeks ago

Fix for ATR-30913: added a callback to outputThread for each noMoreEvent.

Simone Sottocornola authored 4 weeks ago and Jiri Masik committed 4 weeks ago

WebdaqHistSvc: fast publication addition

Stewart Martin-Haugh authored 1 month ago and Andrea Coccaro committed 1 month ago

Fix rewriteLVL1 result - only working with RoIBResult

Will Buttinger authored 1 month ago and Petar Bokan committed 1 month ago

Add declarations of EventInfo creation by eventloop infrastructure

JiveXML: Silence clang warnings

Simone Sottocornola authored 2 months ago and Tim Martin committed 2 months ago

WebdaqHistSvc: Fixed bug in stop

Simone Sottocornola authored 2 months ago and Antonio De Maria committed 2 months ago

WebdaqHistSvc: added use of TDAQ_OH_SERVER variable

Simone Sottocornola authored 2 months ago and Antonio De Maria committed 2 months ago

Added slot and frequency Gaudi option, removed PSC stuff and added partition name from TDAQ_PARTITION variable

Simone Sottocornola authored 3 months ago and Jiri Masik committed 3 months ago

Fix for ATR-30575

HltEventLoopMgr: fix race condition during stop

See merge request atlas/athena!75801

(cherry picked from commit 17cad1ef)

5550169e HltEventLoopMgr: fix race condition during stop

Co-authored-by: Vakhtang Tsulaia <vakhtang.tsulaia@cern.ch>

Let Ctrl-C (SIGINT) be handled by the HLTMPPU to allow for a proper
shutdown and cleanup.

Relates to ATR-30452.

Deleting the `HLT::LoopThread` objects is not thread-safe and can lead
to race conditions between the main thread deleting them and the
callbacks still using them. Also `std::unique_ptr::reset` is not atomic.
Instead of deleting,  just wait for the threads to finish. For
stop/start transitions, the objects will be replaced during the next call
to `nextEvent`.

Relates to ATR-30102.

Make the ByteStreamCnvSvc a component-only library and move the
remaining public headers to ByteStreamCnvSvcBase.

Although we do not support bytestream creation in TrigByteStreamCnvSvc,
we have to at least provide a dummy interface implementation because
converters retrieve the ByteStreamCnvSvc during initialize even though
it's only needed within `createRep`.

Move the `IByteStreamEventAccess` interface inheritance to the derived
classes and use auto-declaring interface syntax.

Stewart Martin-Haugh authored 6 months ago and Duc Ta committed 6 months ago

Remove ROBDataProviderSvc EventContext fallbacks and adapt clients

Frank Winklmeier authored 6 months ago and Johannes Elmsheuser committed 6 months ago

Update to Gaudi v39r0

Add link flags to suppress maybe-uninitialized warnings seen in the LTO
build originating from the boost parser code.

Tim Martin authored 8 months ago and Benjamin Philip Kerridge committed 8 months ago

[ATR-28536][ATR-28717][ATR-28782] Working towards a trigger EDM refactor and integrating R4 into the EDM (part B)

[ATR-28536][ATR-28717][ATR-28782] Working towards a trigger EDM refactor and integrating R4 into the EDM (part B)

Remove dependency on the `OWLTime` class from tdaq in preparation for
Phase-II code updates.

Remove unused code and migrate to `std::format`.

Jennifer Rachel Curran authored 9 months ago and Sara Alderweireldt committed 9 months ago

increase string length in root files (truncations occuring in ART tests - also could be needed for HI runs)

Remove the unused DPD steps and general package cleanup.

Closes ATR-27802.

Delay enabling of interactive mode after the python configuration.

Fixes ATR-29543.

Follow-up to !71038 removing more of the now code for running legacy
configurations.

Now that the HltAsyncEventLoopMgr is used by default we can drop the old
one and restore the original naming.

Andrii Verbytskyi authored 10 months ago and Frank Winklmeier committed 10 months ago

Use  string::ends_with() from STL and not from boost

Need to set the `create=True` on `CoreDumpSvc` to ensure it is always
created even if no client asks for it. Fixes missing signal handling in HLT.

Closes ATR-29451.

After !71207, the transform code no longer appends an explicit `--CA` to
the athena command line.

Closes ATR-29357.

Dump the scheduler state on soft timeout so we can identify the eventID
of the slow event. Requires gaudi/Gaudi!1579 to actually have any effect.

Relates to ATR-27697.

Jennifer Rachel Curran authored 11 months ago and Petar Bokan committed 11 months ago

Reduce error to warning for empty file production and check that the stream name does not get double counted in trigbs_extractStream.py leading to duplicated events being written out