Peter Doak

I don't think banning start/stop is the right solution. I hadn't previously noticed but the scoped ScopedTimer + NewTimer + TimerManager design just a problem with exceptions. It's quite likely...

2b seems like a good idea since this should be a development time error, i.e. a formal release of the code should not be able to produce it, input that...

Yes I do. But doxygen knows nothing about them not rendering because they are just latex thrown into comments without the correct markup. See link above. In general everyone who...

This plan see a good intial breakup. A few things I think will make this more successful. - Make clear distinctions between unit and integration tests as the plan progresses....

This is sort of correct as of clangs docs ``` -Wliteral-range This diagnostic is enabled by default. Diagnostic text: warning: floating-point comparison is always | true false | true |...

This was a prexisting thing with species set so I used it to avoid constructor time dependence on ParticleSet. But it was never a good smell. Need a `std::vector getGroupSizes()...

That was my thought otherwise that code is going to end up everywhere I need that vector on construction and don't want vague ParticleSet constructor argument.

`get_group_sizes` would be the accessor getGroupSizes indicates it could be anything. But since it returns by value and not reference perhaps buildGroupSizeVec is better.

Named/Tagged arguments is something python got right. So should we: - [ ] rename `qmc_run_and_check_custom_scalar` to something like `qmc_system_test` - [ ] Add named argument to `qmc_system_test` to support missing...

I think that the global operations belong in a separate interface. The SPOSet interface should not be both a walker/crowd multiplicity interface and an interface to a global set of...