When we print the list of accessed branches, we were sorting by number
of entries read.  But if two branches were tied, then the result of the
sort was not determined.  Change to compare by branch name if the entry
count is tied.

Infinite recursion.

This is for ATEAM-719
std::pairs coming from xAOD::ByteStreamAuxContainer_v1 std::maps

This was already committed in 21.2; adding it here as pointed out in
ATLASG-1502 (I didn't realize it was missing, if I'm honest).

Basically, the truth link may have a different decoration name when
using some derivations. This lets the helper fall back to the
alternative name if the retrieval of the initial decoration fails.

Allow overriding the default conditions tag when running sg-dump.py.

Frank Berghaus authored 4 years ago and Vakhtang Tsulaia committed 4 years ago

Recusing the seek function had the effect of calling the findEvent a
second time. As findEvent remembers its work the cost is low, but should
only return the same result as before.

If an event is rejected for an output stream, there won't be a DataHeader
made for that stream.  In EventFormatStreamHelperTool, handle the case
of a missing DataHeader.

Fix for the case where xAODMuon doesn't exist.

Attila Krasznahorkay authored 4 years ago and Adam Edward Barton committed 4 years ago

It was introduced in ROOT 6.23/02. We don't make any use of it at the
moment, but maybe later on...

At the same time removed a ROOT 5/6 check from the code. :-P

Frank Berghaus authored 4 years ago and Adam Edward Barton committed 4 years ago

Also delete raw memory buffer read after the event has been build from
it. In case an exception interrupts event building still erase the raw
buffer.

Aaron James Armbruster authored 4 years ago and Adam Edward Barton committed 4 years ago

update format of overlap xml reader to latest; move TCG/RPCRecRoiSvc to switch between run2/3 data formats

Need to load a few more dictionaries early, in order to avoid some
cling assertion failures.

Frank Berghaus authored 4 years ago and Adam Edward Barton committed 4 years ago

Correct the type names and keys the algorithm will write to the output.

Scott Snyder authored 4 years ago and Adam Edward Barton committed 4 years ago

In some cases, some dictionary load requests will fail if they are done
in the `wrong' order.  This is a cling bug (as of root 6.20.00), but
we can work around it here.  We add a TClassGenerator so that if,
for example, we're trying to load ElementLink<T>, we first load T.
And similarly for a few other Athena template classes.
This is observed to be sufficient to fix the order-dependent failures.
See ATR-22681.

Also instantiate the BaseInfo specializations for containers in dictionaries.
This will be needed to work around other cling bugs which are causing
assertion failures in the dbg build.

The EventStorage library assumed that the client takes ownership of the
content read in from disk. The buffer is no longer needed after the
event is built, so delete it afterwards. In case an exception is raised
during the event building delete the buffer before re-raising the
exception.

Add FlowElementContainer to the list of types we can dump.

Hello,

This MR is fixing a bug in a TauDefs enum.
The RNN electron veto working points were recently introduced in the IsTauFlag enum, at positions >=32.
However, we encode IsTauFlag on a bitset of 32 bits:
https://gitlab.cern.ch/atlas/athena/-/blob/master/Event/xAOD/xAODTau/Root/TauJet_v3.cxx#L274-287
So we cannot use anything higher than 31.
Consequently, I have moved the EleRNN WPs lower in the enum.

Currently, if we set EleRNNLoose=1, we read EleRNNLoose=0.
With the bug fix, the WPs work as intended.

The only change in the reconstruction output is the value of "isTauFlags" that encodes the tau WPs.
The RNN eVeto is new in R22, so there is no concern with backward compatibility.

Cheers,
Bertrand

Nils Krumnack authored 4 years ago and Nils Erik Krumnack committed 4 years ago

This is pretty much a 1-to-1 copy of the packages as is in 21.2 (plus
two enums in xAODPrimitives), followed by replacing the use of TAccept
with AcceptData+AcceptInfo.

Need to	call currentContext() before starting leakcheck.

Need to	call currentContext() before starting leakcheck.

Tim Martin authored 4 years ago and Walter Lampl committed 4 years ago

Add 'mayBeRemapped' boolean to TrigComposite link fetching interface, default FALSE (to save CPU). Add link deletion functions.

Bertrand Martin Dit Latour authored 4 years ago and Walter Lampl committed 4 years ago

Hello,

This MR is adding a tau track flag to identify large radius tracks more easily, when LRTs are enabled in the tau reconstruction.
We can now retrieve specifically the LRTs associated with the tau via tau->tracks(xAOD::TauJetParameters::TauTrackFlag::LargeRadiusTrack).
This flag is agnostic w.r.t. the track classification, it is only used to tag LRTs, which will be useful e.g. to count LRTs within 0.4 of the tau axis.

As LRTs are not used by default, this MR does not change the reconstruction output.

Cheers,
Bertrand