This MR in another attempt to reduce the number of differences between CA/old style. It should have NO impact on production, since it is only touching CA-configuration, which is not yet run by default.
In my local tests (using test_trf_q445_r2a_ca_mt_compare.sh) it changes the totals from the nightly values of
nMuons: 2 events (out of 100) muon1pt: 54 events (out of 100) muon1eta: 4 events (out of 100) muon1phi: 10 events (out of 100)
Summary of differences: nMuons: 2 events (out of 100) muon1pt: 15 events (out of 100) muon1eta: 5 events (out of 100) muon1phi: 5 events (out of 100)
To find these problems I hacked ToolSvc in Gaudi to warn about unconfigured tools with non-default names, since for CA config this can mean that, rather being configured from a public tool as in old-style, instead you get an unconfigued tool. So it's a serious category of mis-configuration. Some of the changes are to remove some of the worst false positives (there are cases where a simple private tool really doesn't need configuration ... IMO we should just use the default name in such cases, but I don't want CA and old-style naming differences until we have this migration finished).
I also merged with !54522 (merged) where @tstreble changed the names of some tools. What he did is definitely more correct, but since I still need to compare old/CA I have introduced a variable (
use_old_names) which I can set to True in my local tests to revert to the old-style naming (I will probably need to extend this to other pieces of configuration in a later MR). Later on, this variable and associated configuration will be removed.
In addition to the
test_trf_q445_r2a_ca_mt_compare.sh mentioned above, I have checked this locally with
Please make sure this is squashed! Rebasing was tricky to do, given the number of clashes, and so I'm being lazy.
Please merge after !54872 (merged) since this includes that commit.