configuring logging is awkward and unreliable

[Alan Griffiths]

We have three(!) ways to configure logging:

1. (On startup configuring the server) server.override_the_logger([&]() ...
2. (On startup configuring the client by a debug API) Override ConnectionConfiguration::the_logger()
3. (At any time) mir::logging::set_logger(config->the_logger());

All of these are C++ APIs and none are part of the client "toolkit".

The first method is a public API and directs the output of the server reports and of calls to mir::log* (at least until the third method is used).

The second method is a private API and directs the output of client reports. (But not calls to mir::log*.)

The third method is a public C++ API and directs the output of calls to mir::log* and the client reports (unless the second method was used). (But not the server reports.)

It wasn't confusing before the third method was introduced (with the mir::log* functions). Maybe we should deprecate it?

[Chris Halse Rogers (raof)]

Also, *none* of these ways are available to clients.

I think we should probably scrap (3), not use (2), and instead provide an actual client API for configuring the client logger.

[Mir CI Bot (mir-ci-bot)]

Fix committed into lp:mir at revision None, scheduled for release in mir, milestone 0.21.0

[Alan Griffiths (alan-griffiths)]

The branch that landed changed some behaviour without making the result any less confusing

[Michał Sawicz (saviq)]

Syncing task from Mir.