@karalekas how do you feel about merging this? kyle has some interest in a cousin PR, #874, which has some friction with this one. it would be good to either get in this first or decide to drop it.

Is it possible to specify which memory regions should be retrieved from the QVM? As long as the parameter classical_addresses of _qvm_run() in https://github.com/rigetti/pyquil/pull/873/files#diff-a58f887d22e0687c320a6c4ddf367130L476 is set using get_classical_addresses_from_program, will the QVM still return memory regions other than "ro"?

(Thanks for your PR, I needed to read classical memory beyond "ro" as well -- your work was a great starting point!)

My understanding is that the classical_addresses parameter (or, rather, the analogue of that parameter that actually appears in the QVM payload; I'm not sure if the python parameter gets substantially mangled before making it to the payload) is a dictionary mapping memory names to either a list of indices that the QVM will report on or to True, a synonym for 'all indices'. If you omit a key from this dictionary, that entire memory region won't be reported to the user after a run. If you convert True to some other list of addresses, those addresses not appearing in the list won't be reported to the user after a run.

I'm pretty sure that get_classical_addresses_from_program is set up to have the QVM report everything, and that pyQuil master (but not on this branch) is set up to discard all regions but the ro region.

Also, I deleted the changes.rst entry when resolving the merge conflict (on purpose). Please add the entry back, but now to CHANGELOG.md:

• Add the ability to query for other memory regions after both QPU and QVM runs.
  This removes a previously unnecessary restriction on the QVM, although ro
  remains the only QPU-writeable memory region during Quil execution (@ecpeterson, Enable more memory regions on the QVM #873).

OK: I replied to the sets of comments from @karalekas and from @appleby .