This is the last major component of the quake utilities to be released. To
be honest, I have been a little reticent to release this because most of the
actual qc code is basically rather embarassing crap. The time never became
available to even give it a good top to bottom going over. I never spent
any quality engineering time on my parts, American wrote a lot of qc code,
and even Romero has a bit of work in there. It is a mess. If you look
through the code and occasionally think "This is stupid!", you are probably
right...
The compiler itself can be drastically sped up by just replacing the symbol
searches with binary trees or hashing. We remotely compile on our alpha, so
it hasn't been a big enough issue for me to do it, but as the code size
grows and grows it will be done sooner or later.
The resulting code is horribly nieve and space ineficient (twleve bytes /
instruction). If common subexpression removal was added, the instruction
count could probably be cut nearly in half. I would have liked to have done
a better job at this, but this was my first compiler front end, and I had a
ton of other things fighting for my time. The next one will turn out
better. (wow, I'm making a lot of excuses here, aren't I?)
Qcc also performs some other maintenence functions for us, like rebuildinng
all the brush models and making pak files, but those functions are only
usefull if you have created all new data for everything. models.qc and
sprites.qc don't actually generate any code, they are just parsed by
modelgen and spritegen and included for completeness.
To modify the quake program code, set up a new game directory parallel with
id1, and containing a "progs" subdirectory. Copy all the .qc files and
progs.src into that, and just run qcc from that directory. That will
compile all of the files listed in progs.src and (if there aren't any
errors) generate a new progs.dat file in the parent directory.
As a simple test, open the client.qc file, go to the ClientObituary function
at the end, and change some of the messages.
The directory structure will look something like:
/quake/quake.exe
/quake/id1/
/quake/mygame/progs.dat
/quake/mygame/progs/progs.src
/quake/mygame/progs/world.qc
/quake/mygame/progs/client.qc
/quake/mygame/progs/... etc ...
Run quake with "-game mygame", which will cause quake to look for data in
the mygame directory before falling back to id1. In this example, it will
find the new progs.dat from mygame, and take everything else from id1. You
can type "path" at the quake console to verify the current search order of
directories and pak files. THIS WILL ONLY WORK WITH A REGISTERED VERSION OF
QUAKE.
The header qcc.h has the language spec and some documentation, but I'm not
positive if it is all current.
The only documentation for the various builtin functions I can offer is the
source code used by quake. See builtin.c. Some of them are required to do
things outside the scope of the qc world, and some are just there for speed
reasons.
PLEASE don't ask me questions about all this!
John Carmack
