History of HCC & TOP Amiga Versions:
------------------------------------

The first time I got the compiler was 'Sozobon-C' from Fish #171. Since I
always wanted a C compiler (Until now HCC is my first and only C compiler), I
looked what it would do, but it seemed not to be a working tool. So I put it
back to the disk box.

The next version was 'Zc' from Fish #193. This was a modified version by Joe
Montgomery. The output of this program looked as if A68k would understand this,
and so I decided to do the simplest test for a C compiler written in C: I tried
to let the compiler compile itself. This was my fault (but your luck), since
several weeks of error-detection followed. Joe Montgomery did some serious
mistakes in the Motorola conversion routines. Every output file of Zc had to be
edited manually. The linker missed many names that should be XDEF'd in other
program parts. Then, when I linked the first version of Zc that was fully
XDEF'd, the linker missed something like 'getc', 'printf' etc. I found out that
this are standard C library functions. I searched for this functions and found
them in the 'dLibs' PD standard C library by Dale Schumacher, which was
released with the first version of Sozobon-C on Fish #171. So I tried to
compile this functions. OK, naturally, they needed other library functions of a
lower level, which needed other functions... at the lowest level, there were
system calls to TOS and GEMDOS, since the library was designed for the Atari
ST. But the machine in front of me looked different... what should I do? I
decided to replace the Atari system calls with equivalent calls to AmigaDOS,
compiled and linked all together, and... got a guru, then another guru, etc.
But only a few weeks later, it was done: the compiler could compile itself, but
only when his output was manually edited, and only when some delicate things in
the source were changed (for example, every constant number had to be
explicitly casted to int!).

The next step was a version that fully automatical compiled itself without
manual editing. I decided not to release this version, since my self-written
library routines were VERY slow. It took several hours to compile the complete
compiler. Besides, I wanted a more complete standard library.

When searching through the Fish disks, I found Fish #260 with CClib.library by
Robert Albrecht. This looked to be the solution for my problem. When reading
through the manual, I found that this library expected 32 bit int's, but my
compiler used 16 bit int's. OK, the flag -L was created to force int's to 32
bit. Then the startup code was modified. CClib.library had to be called via
JMP, not via JSR, so special interface stubs had to be created. Then everything
was linked together, started and... crashed. When debugging, I found that the
library destroyed the contents of some registers which were used by the
compiler to hold register variables. I forbid the compiler to use them, then it
worked (and the compiler was faster and shorter!).

Before releasing this compiler, I decided to report the found bugs to Sozobon,
Limited, and to ask about a newer version. I send my bug reports, US$10, and an
offer to create a compiler that would work on both systems via #ifdef AMIGA,
#ifdef ATARI_ST. I received version 1.2 of the compiler, optimizer, etc. on two
Atari-formatted disks, but no answer. After solving the problem of reading
Atari disks, I updated the compiler and optimizer. This time I implementd the
conversion to Motorola format into the optimizer. Nearly every change to the
original is commented (Try 'Search #?.c search TetiSoft').

When inspecting Fish #314 with Zc, I found that this compiler could not compile
itself. The changes made to the source were mostly not documented. The only
things I took from this version are Jeff Lydiatt's math library routines and
his implementation of (void *) as universal pointer (and, of course, the latest
version of A68k).

The next thing was to test the compiler with other programs than itself. I
found that he could compile A68k. OK. But the 'CManual' from Fish #337 made
problems. The compiler did not distinguish between 'unsigned' and 'unsigned
short'! After fixing this problem (and some others), HCC could compile most of
the 'CManual' examples.

Next thing in this history is the successfull implementation of CClib.library
V3.0, which allowed to use D3 again for register variables, and had the
bugs fixed I reported to Robert W. Albrecht.

Wrote new startup that alerts with a text message if CClib.library V3 or
higher not found.

The latest enhancement is the implementation of the 2.0 Atari version I
received from Sozobon, Limited.

OK, and now I think it's time to release the compiler. Let's make the first
entry in the table:

May 1991:
Amiga version 1.1. First release, portation of Atari version 2.0.


Greetings to all AMIGA freaks,

Detlef Wuerkner
Werrastrasse 7
D-6300 Giessen
GERMANY