cmssw: Enforce that Process can have at most one Schedule object, and the label must be schedule
There should be 0 or 1 cms.Schedule
objects in the cms.Process
, and the label of the cms.Schedule
should be schedule
.
Issue https://github.com/cms-sw/cmssw/issues/35624 made it apparent that ConfigBuilder
and HLT menus have been using cms.Schedule
object labelled HLTSchedule
for a long time. We should look into expressing the desired behavior with the label process.schedule
, after which we could enforce the rule.
About this issue
- Original URL
- State: open
- Created 3 years ago
- Comments: 16 (16 by maintainers)
Commits related to this issue
- make HLT use "schedule" For more information, see https://github.com/cms-sw/cmssw/issues/35842 — committed to missirol/cmssw by missirol 3 years ago
- make HLT use "schedule" For more information, see https://github.com/cms-sw/cmssw/issues/35842 — committed to silviodonato/cmssw by missirol 3 years ago
Hi all, a quick update on this issue from the HLT side.
A reminder: the goal is to deploy an update of
ConfDB
(external to CMSSW) so that it outputsschedule
instead ofHLTSchedule
(thus removing all refs toHLTSchedule
inmaster
moving forward).This
ConfDB
update can have side effects for users of12_{0,1}_X
, so HLT needs to make sure this is minimised (users of releases “< 12_X_Y” have access only to an older version of ConfDB, a.k.a. “ConfDB-v2”, which will not be updated anyway).As discussed in #36237, we converged on the following plan:
12_0_X
: needs #36251, which is an ad-hoc change for this release, so that HLT can continue to deliverHLTSchedule
to its clients.12_1_X
: needs #36252, which is a backport of #35858 (and a related PR), putting12_1_X
on the same footing as12_2_X
; since HLT development is mostly done in12_1
and12_2
these days, making the two branches as close as possible is considered the preferred way to go.12_2_X
(currently,master
): ready for theConfDB
update, no further action needed.This plan implies that a new
12_0
release and a new12_1
release will be needed. After that, it would be possible to deploy theConfDB
update, and complete the HLT side of this issue.Please let us know if you see any issues with this plan.
Thanks @missirol!
We could also do it for 12_2_0 already, ie, between now and early January and see during this period if there are unexpected side effects. The ConfDb change only affects the extracted py configs. In any case we need to advertise in Monday/Wednesday TSG meetings
@missirol I can’t comment when would be a good time to make the change in ConfDB itself (except I’m fully with you that it needs to be planned carefully), but I thank you for all the work in CMSSW side!