[BusyBox] supported targets

[Larry Doolittle]

Matt wrote:

> Please do not take this as a slight against you, Larry ...

No, I understand completely.  The patch did its job -- it clarified
the discussion.  Plus, if someone jumps out of the woodwork next
week and says "BusyBox 0.49 doesn't work on my uClinux-powered
toilet paper dispenser because my stack is too small" we know
what to tell him.  (Starting with, "why didn't you mention this
last week ...")

>    arches: x86, ARM, MIPS, SH3/4 (I think)
>    libcs: libc5, glibc, uClibc
>    kernels: Linux 2.0, 2.2, 2.4 (unless someone says otherwise)

That's a good start.  I think this should be documented.
Let me add a few details and questions:

archs:  most code is architecture neutral.  The exceptions are:
      insmod:  only x86, ARM, SH3/4
      mkswap:  anything supported by Linux 2.2.0pre9

libcs: libc5, glibc, uClibc (what about newLib or diet libc?)

kernels: Stock Linux 2.0, 2.2, 2.4 (what about uCLinux 2.0, 2.4?)

It looks like there is residual arch specific code left over from
the sources from which BusyBox was assembled, including special
treatment of sparc and alpha.

Who is the "customer" for the fake_sparc handling in uname.c?
Can and should that code be generalized?

When was the last time the __sparc__ code in init.c and loadfont.c
was tested?  When was the last time the __alpha__ code in dmesg.c
and syslogd.c was tested?

Is #if #cpu(sparc) really the right test for choosing old termio?
(cmdedit.c, more.c)

After these questions are clarified, I propose including some of
this material in the README.

      - Larry