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

The two classes are functionally quite identical, so this is mostly a
straightforward string replacement, with a couple of manual fixes.

Since Gaudi's StopperAlg is not available in the Athena runtime anymore.

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

This has been a long time coming, but one of the reasons for having
AsgMessaging as a separate package from AsgTools is to allow
xAODRootAccess to depend on it and use the same status codes and
messaging as the rest of analysis base.  This should take part of the
former.

I had to add the RECOVERABLE code to the main status code, but since
that is also part of the Gaudi StatusCode that seems like a reasonable
change/extension.

Apart from that this is mostly a global search replace to move all
users from TReturnCode to StatusCode, but apart from fixing up a
couple of places related to the status code itself I did not try to
fix/improve any code in the process.

Note that this also revealed a number of tests that failed because
they used RETURN_CHECK() incorrectly.  This commit doesn't try to
address that.

Remove `atlas_depends_on_subdirs` and update link dependencies to what
is actually used by the source files.

TAG files are no longer being produced. Delete this obsolete e/gamma TAG
maker and code referencing it.

EGamma ambiguity. Add a few comments, format a bit, use override final, default dtor, removed unneeded finalize

This is superseced by AsgMessaging/AsgMessaging*.h, though in some
places extra includes are needed.

Those have been superseded by the files in AsgMessaging.

That has been superseded by AsgMessaging/StatusCode.h.

That has been superseded by AsgMessaging/MessageCheck.h.

Since the package was not built as part of Athena before, it is now
added to that project as well.

The test in Athena doesn't actually run anything... It sets up a job,
but doesn't run any event loop. However I didn't want to completely
re-design the test to set things up correctly for an Athena test.
(The Athena and standalone tests should just be done using separate
Python files, as they don't actually share any meaningful amount of
code with each other.)

Just loading all the dictionaries that @akrasna had in his
LoadDictionaries function in 21.2.

There were some lines in there that seemed to be targeted at older
versions of root.

We somehow ended up with two copies of the same interface class.  The
problem with having two interfaces for the same task was that tools
implementing the different interfaces could not be used
interchangeably.  That specifically broke the configuration of some of
the common CP algorithm sequences, which use the same algorithm class
for all selection tools.

I was somewhat torn between which of the two classes to keep, and
while I originally wanted to keep `CP::ISelectionTool`, it turned out
that `IAsgSelectionTool` was much easier to keep.  That is because the
`AcceptInfo` and `AcceptData` classes are in the same package as the
interface class, and anybody who wants to use or implement that
interface will need those two classes as well.

One could potentially move all three classes together to
AsgAnalysisInterfaces if that is considered advantageous, but that
would be a bigger undertaking and they would all have to be moved
together.

These tests were only run in AnalysisBase, however they have been
failing since including this package in master, so for now I am
disabling them, leaving it to the domain experts to fix them and
reenable them.