Path: utzoo!attcan!uunet!van-bc!root
From: bhenning@bhami.vnet.van-bc.UUCP (Bill Henning)
Newsgroups: comp.sys.amiga
Subject: Lattice C V5.0
Message-ID: <2119@van-bc.UUCP>
Date: 8 Jan 89 03:35:38 GMT
Sender: root@van-bc.UUCP
Lines: 63


   I bought Lattice C V5.0 last thursday, and I thought I would relate some of
my experiences with it. Overall, I am impressed, and I do like it. However...

   After painlessly installing the package on my hard drive, I thought I would
compile  some  C  programs...   this  is  when the trouble started.  One of my
previous  projects has been a cross-assembler for a microcontroller.  Three of
the five modules produced a CXERR 26 in line 0 from LC2, which is a 'Temporary
save  or  restore error'.  My guess is that the peep hole optimizer is getting
confused and removing the store of p's new value back into its variable.

  After  considerable  (and  judicious)  use of '#ifdef FINDBUG...#endif" in a
binary  search on the smallest of the files I was able to isolate the problem.
At  the  end  of  this  message  I've  included  a  small  code  fragment that
illustrates  the  problem.   LC2  cannot handle *p++ where p is declared as an
extern pointer to char.  This is a serious and terrible bug.  C code uses *p++
all  over  the  place.   Fortunately,  most cases use it within the scope of a
single  module or function (but my assembler used it across module boundaries,
with  a  .h  file declaring it as an extern).  I tried -b0, no dice.  The only
solution  I  found  was  to  separate  it into two statements, and casting the
pointer to a long before incrementing it.

   Before anyone thinks that I'm totally disgruntled with the compiler I
should note the following:

   - the debugger is absolutely fantastic. Lattice deserves a pat on the
     back for it.

   - The inclusion of a profiler, lmk, and other utilities is commendable

   - The manuals are excellent. Nicely layed out, nicely indexed. I am
     also grateful for the library function compatibility/origin list

Great work John Toebes (sp?) and others! Now if only LC2 were fixed...

Here is the code illustrating the bug and workaround:

extern char ch;
extern char far *p;
#define BUG
void dummy() {
#ifdef BUG
	ch = *p++;
#else   /* and a temporary work-around
	ch = *p;
	(long)p++;
#endif
}

   A  couple of other minor gripes.  When compiling some source with -b0 Blink
complained that it could not find cxovf and cxbrk.  Fortunately the sources to
those  functions were included, so it was trivial to compile them and add them
to lc.lib.  Where is grep.lib?


--
##############################################################################
# Amiga // #   UUCP: {alberta!ubc-vision,ubc-cs,uunet}!van-bc!bhami!bhenning #
# 1000 //  #   -or-  bhenning@bhami.wimsey.bc.ca /* untested at this time */ #
#     //   #   -or-  bhenning@bhami.van-bc.UUCP  /* I prefer UUCP mail    */ #
# \\ //    # BITNET: USERCCPM@SFU.BITNET /* I rarely check for other mail */ #
#  \X/bhami#   ARPA: William_Henning%sfu@um.cc.umich.edu                     #
##############################################################################