[MERGE][#248153] Add autopack RPC

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Martin Pool wrote:
> Martin Pool has voted comment.
> Status is now: Semi-approved
> Comment:
> This definitely deserves a mention in NEWS, including some kind of
> quantification of what improvement it will give you, and a description of
> what will happen if the server is too old.
> 
> +    def autopack(self):
> +        path = self.bzrdir._path_for_remote_call(self._client)
> +        try:
> +            response = self._client.call('PackRepository.autopack', path)
> +        except errors.ErrorFromSmartServer, err:
> +            self._translate_error(err)
> +        except errors.UnknownSmartMethod:
> +            self._ensure_real()
> +            self._real_repository._pack_collection.autopack()
> +            return
> +        if self._real_repository is not None:
> +            # Reset the real repository's cache of pack names.
> +            self._real_repository._pack_collection.reload_pack_names()
> +        if response != ('ok',):
> +            raise errors.UnexpectedSmartServerResponse(response)
> +
> 
> While we're here, it seems like an obvious opportunity to have the
> PackRepository.autopack command return the pack names data, so that we
> don't need to do another roundtrip.  I can see two counterarguments
> though, one being that it might not be trivial to inject it into the
> _real_repository, and the other that if we can often get away without
> vfs-based operations it may be unnecessary.  But at least the second does
> not seem to be true in the near future.

I agree that returning 'pack-names' seems reasonable, and it shouldn't
be very hard to factor 'reload_pack_names' one more time to allow
passing in the name list.

John
=:->

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkkJvvkACgkQJdeBCYSNAANCoQCgpaFDincHAC9W1TwM737qoIX8
ecsAoItT/Sd0Givd/6XIhoCpcQ6dJWqY
=Sp8D
-----END PGP SIGNATURE-----