Remove support for MS-DOS

[nbriggs]

This is the first cut at removing support for MS-DOS as I don't think we're ever going to build there again

• remove (with "unifdef") all code depending on symbol DOS
• remove all .c and .h files only relevant to DOS
• remove these from CMake and traditional build
• remove the DOS makefile
• cleanup a few leftover defined(DOS) in logical expressions and fix endif comments
• remove a couple of assembler modules for DOS only

[waywardmonkeys]

One thing that stopped me from doing this before was that I wasn’t sure if some of the code for handling drives in oaths might be useful for Windows ... thoughts?

[nbriggs]

Yeah, you have a point there. I wish we could dispense with drive letters, though. Some of the ugliest code is in there.

[nbriggs]

I'll take a shot at revising this to not strip out the file system related code.
I'm not sure what you envision on the input side if your new display code is implemented -- even if it's an X-like input there'll be a whole lot of specific code to deal with a Windows (rather than DOS) keyboard/mouse input.

[waywardmonkeys]

For my display work, I'm using Sokol (https://github.com/floooh/sokol). Specifically, I am using sokol_gfx.h and parts of sokol_app.h (unless I can rework some other things and then I can use it directly).

This lets us run the same code on Unix, macOS, Windows, Web, and more.

I don't worry a lot about the existing code for handling keyboard and mouse input. :)

[masinter]

My thoughts have involved a URL-based file system, and handling drive letters the way whatwg/url does.

[nbriggs]

Cool. Using the sokol_app "event-based keyboard, mouse and touch input" I guess. It looks as though it will have issues of similar complexity to the way the X input events needed to be handled.

[masinter]

line 61 of (new) src/ether.c has #endif /* DOS */ but the ifdef remover removed the wrong one: #endif /* MAIKO_ENABLE_INTERNET */

[masinter]

I think this might gate some other work

[nbriggs]

@masinter -- I'll fix the endif problem. I'm probably going to revise this substantially because Bruce asked that the file-system code handling drive letters be left for when/if he has a native Windows port. It also has conflicting commits with the current master so I'll be needing to rebase and cleanup before merging.