GitLab

Use GNU indent to process the source into a consistent pattern:

`-nut -bli0 -i4 -l120 -lp -as -nbbo -cdw -npsl -npcs -cli4`

`ADD rx, pc, #y` (and SUB) is processed into ADR statements instead.
The JFTest tool now outputs differences as context diffs instead of
the whole output.

The suffix ,ffa is now recognised as a module and gives the base
PC address as 0, rather than 0x800.

Because of the way that file handles were not 'eof' until a failed read
had occurred, we were reading one word too many. This has been corrected,
and the correct number of words are read (rounded up to a word).

Using '"string" + 1' as a means of referencing different parts of a static
string isn't really that good a practice, efficient though it may be in some
cases. It's still not 'fixed', but the warnings have been quoshed by using
&"string"[1] instead.

The 'use' cannot be used inside an eval - it won't trap. So instead,
we just do a 'require'.

A basic test that we can disassemble a module - the old LegacyBBC module
that provides a bunch of the old BBC interfaces.

A shift left of 32 is undefined behaviour for a 32bit type. The result
in the MacOS compiled binary was that it would shift by 0. Because then
the result was added to the shift right, the result was double that
expected. Fortunately, this is an easy fix.

Generated Makefile will let us build this for Unix-isms.

In order to allow the building of the tool on non-RISC OS systems,
the source files containing the SWI tables are now checked in.

  Initial commit of the ARM disassembler.
Detail:
  * ARMDiss was the initial part of my university project (a Pascal
    decompiler). It can disassemble simple ARM, and compiles under RISC OS
    or unix (IIRC!).
Admin:
  Untested.
Tag:
  ARMDiss-1_01

linking. Moved itoa and padwithspaces to utils.c.

alongside Debugger. Should be portable now.

LDR/STR, LDM/STM, B[L].
Stub functions report co-processor unsupported.