(Cannot find an old ticket for this :( thought there was one...)
Fabric 1 had with cd: and with lcd: and similar dual APIs for users that needed to track context-y state (cwd, prefixes list etc) differently for the remote vs local ends.
Invoke only has the one local context and doesn't care.
Fabric 2 inherits from Invoke but failed to grow distinct remote-vs-local APIs before 2.0 (besides the much earlier implementation of "move Invoke's run to Fabric 2's local", which presents its own set of problems around "what kind of Context am I dealing with right now and what methods does it have?").
This needs solving and I assume in the obvious fashion, i.e. growing new API members (named explicitly either after local or remote) and the appropriate state storage (on Connection or in config).
This will sadly be a backwards compatibility break (insofar as with c.cd() currently changes both ends and will revert to changing only one) but the software's only been out a week and I think we can argue that this was a bug. Otherwise it'll be the shortest lived major version number ever 😐
This is strongly related to #1686 which is for the run-vs-local config divergence.
(Cannot find an old ticket for this :( thought there was one...)
Fabric 1 had
with cd:andwith lcd:and similar dual APIs for users that needed to track context-y state (cwd, prefixes list etc) differently for the remote vs local ends.Invoke only has the one local context and doesn't care.
Fabric 2 inherits from Invoke but failed to grow distinct remote-vs-local APIs before 2.0 (besides the much earlier implementation of "move Invoke's
runto Fabric 2'slocal", which presents its own set of problems around "what kind of Context am I dealing with right now and what methods does it have?").This needs solving and I assume in the obvious fashion, i.e. growing new API members (named explicitly either after local or remote) and the appropriate state storage (on Connection or in config).
This will sadly be a backwards compatibility break (insofar as
with c.cd()currently changes both ends and will revert to changing only one) but the software's only been out a week and I think we can argue that this was a bug. Otherwise it'll be the shortest lived major version number ever 😐This is strongly related to #1686 which is for the run-vs-local config divergence.