Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxn!ihnp4!cuae2!ltuxa!we53!busch!wucs!jane
From: jane@wucs.UUCP (Jane Mack)
Newsgroups: net.micro.mac
Subject: Re: Reviews wanted please
Message-ID: <1430@wucs.UUCP>
Date: Tue, 11-Feb-86 13:22:09 EST
Article-I.D.: wucs.1430
Posted: Tue Feb 11 13:22:09 1986
Date-Received: Fri, 14-Feb-86 01:38:55 EST
References: <396@3comvax.UUCP>
Distribution: net
Organization: Washington U. in St. Louis, CS Dept.
Lines: 30
Summary: comments on Aztec C compiler



I've been using the Aztec system for about a year and I'm pretty happy with it,
except of course that it compiles a whole lot slower than the c compiler on our
vax.  Our source code is over 700 kB spread over 77 modules (some of them
quite small) and our executable code is 220 kB.  We've only had a few 
minor problems.

(1) The most troublesome problem is the code generated when you have a handle 
to a record, and you assign the result of some function to one of the fields 
in that record.  The compiler caches the address of the record structure 
in a register, calls the function, and then calculates the address of the 
particluar field of the record that is to receive the result.  It calculates
this result address using the address that it cached before the function call.
Well, as you can guess, if the function allocates a lot of memory, then
memory will be compacted to make room and the record structure will be 
moved.  Hence, the cached address is garbage.

We think this is a bug.  To get around it, either use a temporary variable
to receive the result of the function, or lock the handle before calling the
function.

(2) Make doesn't work when you have as many modules as we do.

(3) To link with the c and math libraries when your executable code is big 
you have to include them both at the beginning and the end of the ln command.

Overall I think the Aztec system is pretty good. 
good luck.
jane mack