I was originally intending to have act-tester-analyse able to take in multiple plans, but it'll be easier to spin out plan merging into a separate tool and have everything else expect one plan; else, we'd have to change the way we save plans, etc.
The idea here is that you'd be able to take multiple plans and combine them into a single plan, with the following provisos:
Each plan would have to have some sort of unique identifier (probably, to begin with, the position in the input file list);
Each plan would have to agree on a version;
Each plan would have to agree on the set of stages that have been completed (this, or an elaborate scheme of sub-plan accounting for stages);
The resulting plan will have all compiler IDs and subject names prefixed by the plan identifier, possibly in some reversible way.
This would appear as a new tool act-tester-merge that is separate from the director loop.
Why would you want to do this? The main reason is so that we can get timeset metrics aggregated over more than one plan, both in terms of compilers/runs and also in terms of stages.
I was originally intending to have
act-tester-analyseable to take in multiple plans, but it'll be easier to spin out plan merging into a separate tool and have everything else expect one plan; else, we'd have to change the way we save plans, etc.The idea here is that you'd be able to take multiple plans and combine them into a single plan, with the following provisos:
This would appear as a new tool
act-tester-mergethat is separate from the director loop.Why would you want to do this? The main reason is so that we can get timeset metrics aggregated over more than one plan, both in terms of compilers/runs and also in terms of stages.