Hi,
Wouldn't it be useful to have the TSE be accurate for such PVs even if they
don't finish processing until 1 or more events later?
Sure, this would be useful (not for LCLS BSA, which has already marked the old pulse as missing, but perhaps for other tools and other kinds of BSAs).
So the idea is that the TSE field of the record is set to -2 so that when the record is processed, normal record processing won't override what asyn set. But the event code then has to be input to asyn in an alternate way and it has to be changeable on the fly. And the data acquisition end of things does the epicsTimeGetEvent if a non-zero event code is available and then gives that timestamp to asyn device support later on. Would be nice to let TSE be the event code but tell record processing not to use it (ie, by something unique in TSEL that doesn't resolve to a PV).