This continues some of the great work that @benma has been doing.
More work towards a CConnman with no globals, so that we can test instances against themselves.
Also, rather than adding to the existing unused parameters
(threadGroup, scheduler), remove them and their header dependencies.
This also allows the port-less AddLocal overload to be removed.
The global test is removed because it's no longer relevant. Overriding via the
-port config option is tested extensively via python tests.
Grr, serves me right for building without Qt. Will fix.
@jonasschnelli Any suggestion where to stick this? The qt build fails because the global accessor has been removed. I'd hate to add the global back just for qt.
Is it possible to add a QSettings value from gArgs without it ending up stored in the config file?
@theuni: I don't know whats the simples change. You could pass it into a GUI constant holder singleton (or pseudo global). Could be held in guiutil.cpp/h. Or just pass it around (to bitcoingui.cpp, etc., but I guess that will lead to a lot of code)
This seems to be the only option that changes operations on the fly. We could just not do that (setRestartRequired(true)).
Otherwise, a clean option might be to register a callback in qt, called when non-restart settings are changed. Could hook up the dataChanged signal to a slot somewhere closer in qt to where the callback can be passed.
Concept ACK
<!--5fd3d806e98f4a0ca80977bb178665a0-->There hasn't been much activity lately and the patch still needs rebase, so I am closing this for now. Please let me know when you want to continue working on this, so the pull request can be re-opened.
Picked up in "net: remove more CConnman globals (theuni) #14856"