From:	CSBVAX::CSBVAX::MRGATE::"SMTP::CRVAX.SRI.COM::RELAY-INFO-VAX" 17-MAR-1989 22:50
To:	MRGATE::"ARISIA::EVERHART"
Subj:	Re: Compress/Decompress problems


Received: From KL.SRI.COM by CRVAX.SRI.COM with TCP; Thu, 16 MAR 89 14:59:13 PDT
Received: from uunet.UU.NET by KL.SRI.COM with TCP; Thu, 16 Mar 89 14:51:02 PST
Received: from munnari.UUCP by uunet.UU.NET (5.61/1.14) with UUCP 
	id AA12413; Thu, 16 Mar 89 17:51:00 -0500
Received: from levels.sait.oz (via mimir) by munnari.oz with SunIII (5.5)
	id AA02795; Fri, 17 Mar 89 09:31:38 EST
	(from CCDN@levels.sait.oz for uunet!info-vax@kl.sri.com)
Received: from levels.sait.oz by PISA.sait.oz (4.44/SAIT 1.0)
	id AA00746; Thu, 16 Mar 89 17:41:24 CST
Message-Id: <8903160711.AA00746@PISA.sait.oz>
Date: Thu, 16 Mar 89 17:49 +1030
From: DAVID NEWALL <munnari!levels.sait.oz.au!CCDN@uunet.UU.NET>
Subject: Re: Compress/Decompress problems
To: info-vax@kl.sri.com
X-Vms-To: in%"info-vax@kl.sri.com"

Path: levels!ccdn
From: ccdn@levels.sait.edu.au (DAVID NEWALL)
Newsgroups: comp.os.vms
Subject: Re: Compress/Decompress problems
Message-ID: <1762@levels.sait.edu.au>
Date: 16 Mar 89 17:48:01 GMT
References: <8903132157.AA09874@ucbvax.Berkeley.EDU>
Organization: Sth Australian Inst of Technology
Lines: 54

In article <8903132157.AA09874@ucbvax.Berkeley.EDU>, Brian.T.Carcich@CUSPIF.TN.CORNELL.EDU writes:
> I have lzw from kuhub, I am running vms 4.5, and I get the following
> results with both the ftp'ed lzdcmp.exe and a rebuilt one:
>
> The decompression step:
>
>      $ lzd/show=all/exp=unix a.a_z a.a_u
>
>      Decompress didn't finish correctly.
>      Error when decompressing "A.A_Z" to "A.A_U"
>      %NONAME-W-NOMSG, Message number 00000000

I am running VMS 5.0 and get similar results.  With /SHOW=(ALL,DEBUG), I
get the following [extract only]:

        $ decompress ...
        ...
        code_count 850
        9: 00 01 48 72 18 c2 06 17 d4

        Read LZ_CLEAR(256)
        code_count 851
        -1:

        Decompress didn't finish correctly.

I assumed a bug.  In fact, I'm sure of it, because I found a guaranteed
bug in lz.h.  Consider the definition of the FAIL macro:

        #define FAIL(why, status)                    \
                fprintf(stderr, "fatal: %s\n", why); \
                longjmp(failure, status);

Consider how it was used (from lzdcm1.c):

        #if VMS_V4
                ...
        #else
                if ((export & EXPORT_VMS) != 0)
                        FAIL("VMS private mode not supported", EXIT_FAILURE);
        #endif

Obviously wrong, yes?  [Hint: it expands to ``if (...) fprintf(); longjmp();'']
The fix for this bug was simple:  put braces around the expansion for FAIL.
But I figure that where there's one bug, there's maybe another; and that's
why the decompression doesn't complete.

Actually, I wish someone could send us a patch to fix this, because lzw
seems otherwise quite good!

David Newall                     Phone:  +61 8 343 3160
Unix Systems Programmer          Fax:    +61 8 349 6939
Academic Computing Service       E-mail: ccdn@levels.sait.oz.au
SA Institute of Technology       Post:   The Levels, South Australia, 5095