De-duplicate code for TerminalDispatch and AdaptDispatch

[zadjii-msft]

Both these classes are doing a lot of the nearly same work at this point. We should probably not, and have more common implementations somehow.

IIRC, when I was writing it, I think part of the idea was to move ITerminalApi up and out of the Terminal to be shared, then have one Dispatch that calls into the API, then an adapter between that API and the conhost API

This is a super crude diagram

         +------------------+
         |  OutputEngine    |
         +---------+--------+
                   |
         +---------v--------+
         | OutputDispatch   |
         +--------+---------+
                  |
         +--------v---------+
         |   ITerminalApi   |
         +---+------------+-+
             |            |
             |            |
+------------v--+     +---v------------+
|  Terminal     |     |  HostAdapter   |
+---------------+     +-------+--------+
                              |
                      +-------v--------+
                      |   outputStream |
                      |      +         |
                      |      |         |
                      |      v         |
                      |   getset       |
                      +----------------+

In this model, Terminal and HostAdapter both implement ITerminalApi, so there only needs to be one Dispatcher. HostAdapter then calls into the conGetSet API we already have.

This would pretty horribly break tests that test the existing structure :/

I could be horribly off base with this though. I came up with this idea like 9 months ago and never wrote it down.

• Should the bold/bright attribute apply to SGR 38/48 indexed colors? #5384 (xterm indexed color problem)

[miniksa]

This sounds reasonable to me. Doesn't matter if it breaks the test. We've already identified that half of the adapter tests are pretty silly anyway. Probably need to revisit that anyway.

[miniksa]

It looks like the TerminalDispatch methods don't really have tests on them either. So it's probably up to this task to consider testing overall.

[j4james]

My initial thought for this was that we would just drop the TerminalDispatch code and replace it entirely with AdaptDispatch. That already has all of the functionality that TerminalDispatch is trying to replicate, and it actually does have tests in place too.

Then the interface that splits the conhost and Terminal implementations could possibly be based on the current ConGetSet interface. It'll require quite a lot of cleanup before it's really suitable for that task, but I think that's probably the right place for it.