Path: utzoo!censor!geac!torsqnt!lethe!yunexus!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!cs.utexas.edu!sun-barr!olivea!samsung!spool.mu.edu!snorkelwacker.mit.edu!bloom-beacon!eru!kth.se!news
From: wahlsten@elixir.lne.kth.se (Jorgen Wahlsten)
Newsgroups: comp.os.minix
Subject: Re: c68 has a switch bug
Message-ID: <1991Mar1.152858.29005@kth.se>
Date: 1 Mar 91 15:28:58 GMT
References: <1087@mwtech.UUCP>
Organization: Dept. of EE, Royal Institute of Technology, Stockholm, Sweden
Lines: 41

In article <1087@mwtech.UUCP> joerg@mwtech.UUCP (Joerg Werner) writes:
>
>I use c68 all the time and I like it. But after I compiled the
>complete Minix-ST 1.5 operating system, mdb (the Minix debugger)
>doesn't work any longer.
>
>Please have a look at the following programm:
> [...]

The bug is in genstmt.c when generating code for the 'switch'-statement.
The expresseion in switch(expr) is assumed to be unsigned (somehow), so
when you have an short int, 'i' in your example c68 never cast'ed it into a
long before looking up the value (it became 65535L instead of -1).
This only happens when you use more than 7 case-statements, that is, when
c68 generates a long-table with addresses instead of cmp/jmp...

I've made a modif. that (i think) work but i haven't tested it that much,
and i'm not sure my modification respects all there is (Better have CvW
confirm). But for a quick and easy change, use the diff that follows...

Regards,
/wahlsten@lne.kth.se

output from 'diff genstmt.c genstmt.c.org':
---------- Cut here! ----------
362d361
<     enum e_bt	    ctyp;
377d375
<     ctyp = stmt->exp->etype;
411d408
< 	if (size == 1 || size == 2)
413,419d409
< 	    ap = g_cast(ap, ctyp, bt_long, F_DREG | F_VOL);
< #endif
< #ifdef INTEL_386
< 	    ap = g_cast(ap, ctyp, bt_long, F_REG);
< #endif
< /*
< #ifdef MC680X0
431d420
< */