MIDAS bug fixes for assembling Muddle

[larsbrinkhoff]

@atsampson has found a number of problems in MIDAS when using it to assemble Muddle.

Check #665 for gory details.

Also, he's been using MIDAS 323 because 328 and later doesn't work. So see what the difference is and fix it. (73, 289, and 308 also work.)

[larsbrinkhoff]

I think the next version of MIDAS should be 459 rather than incremented for every bugfix or git commit.

[larsbrinkhoff]

Also add the INFO; MIDAS ARCHIV file.

[larsbrinkhoff]

MIDAS 323 can multiply a global symbol with a constant. See e.g. MUDSYS; SAVE 176, ; POINT TO REAL CHANNELS SLOTS.

MIDAS 458 won't do that. It'll complain saying "Externals multiplied".

Checking the diff, it seems the logic at label MULTP1 has been arranged somewhat. The 323 version seems to work, but supposedly the change was made for a reason. Some testcases and careful single-stepping should reveal the intent.

[atsampson]

To summarise the changes we made to MIDAS 323 when we were using it for building Muddle:

• Break .ALSO RET] into two lines, so it would compile with MIDAS 458.
• Add XMOVEI to the symbol table, since the Muddle source used it (it doesn't any more) - 458 has this.
• Set the IRCONT flag for all two-word linker directives so they don't get split across a block boundary - 458 needs this fix.
• Define the right symbol name at the start of a CONSTANTS area - 458 needs this fix.

We're now using MIDAS 73 for Muddle, so none of these are strictly necessary, although it would be nice to include the last two in 459 for posterity.

It may well be that the other older code we're building with 323 would actually build with 73/77, so 323 itself might not be necessary.

[larsbrinkhoff]

Unless 323 is necessary to build 77?

[atsampson]

At present we build 73 using a binary of 77 - neither 73 nor 77 will build with 323.