Reduce calls to mega65_io_enable()

[mlund]

To ensure that MEGA65 expanded features are availbel, mega65_io_enable() is called from many functions, for example memory.c::do_dma() and multiple places in fcio.c, see e.g. the call graph produces by Doxygen:

• Remove calls
• Update function documentation that it's the callers responsibility to call it
• Does this break backwards compatibility?

[lydon42]

To the last point, the following parts of core and utilities do enable mega65_io:

• megflash
• onboarding
• mega65_config (does not use libc)
• fdisk (does not use libc)
• tools ftphelper (does not use libc)
• tools tests (either call the function or do the POKE)

So I gather we are save on that side.

As you have added versions, this is a good change as it reduces repetition, which makes the program slower, so if someone forgot to enable MEGA65 IO at the start of the program, they can fix it after reading the release notes.

Additrional idea: If we add a real startup for MEGA65 C programs to the various compilers, then the startup should do that and the user does not need to remember, right?

[mlund]

Is this habit still the case?

mega65-libc/src/fcio.c

Line 557 in 0f7bdab

mega65_io_enable(); // kernal has the disgusting habit of resetting vic

[lydon42]

good question, lets ask @dansanderson

[dansanderson]

The Compendium warns that DOS operations reset the KEY register. I don't see where this happens in code. (Get_DOS doesn't seem to have anything relevant.) I'm still pretty new to the DOS code.

There are only a few direct references to vic+47 (and none to d024) and they're in respectable places (ioinit, c64beg, reset). I gather that any write to d024 that isn't an enable knock will disable, so maybe there are indirect references. The only place I know that resets all VIC registers is the one in ioinit.

I have tried and failed to set a watchpoint on a VIC register, so I don't know if there's an easy way to interrogate this. Ideas welcome. Let me know how else i can help.