INFO-VAX Fri, 11 Jul 2008 Volume 2008 : Issue 385 Contents: Re: 226 Transfer complete question Re: Changes to the freeware web site Re: Changes to the freeware web site Re: Changes to the freeware web site DCL divide by zero WAS: Symbol Substitution Mystery Re: DECW$SERVER_0 at 100% CPU? Re: MTH$ RTL documentation Re: MTH$ RTL documentation Re: Quasi-push technologies Re: Quasi-push technologies Re: Symbol Substitution Mystery Re: Symbol Substitution Mystery Re: Symbol Substitution Mystery Re: Symbol Substitution Mystery Re: Symbol Substitution Mystery Re: Symbol Substitution Mystery Re: Symbol Substitution Mystery ---------------------------------------------------------------------- Date: Fri, 11 Jul 2008 07:36:26 -0700 (PDT) From: "David P. Murphy" Subject: Re: 226 Transfer complete question Message-ID: <5ca1fc26-5935-4694-b3eb-ca61a9689d19@34g2000hsh.googlegroups.com> On Jul 7, 11:54 pm, s...@antinode.info (Steven M. Schweda) wrote: > Why are you looking at _anything_ after "226"? > Why are you looking at anything after "2"? Good answer. Listen to the man, kiwi. Grok the RFC. Coincidentally I am coding the same capability --- parsing FTP output --- in a korn shell script. It's fairly straight forward although I am angry that the line echoing the filename and the line reporting the size and speed are not prefixed by a reply code. In fact the latter can quite possibly begin with a three digit number followed by a space . . . grrrrrrrrrrr! ok dpm ------------------------------ Date: Fri, 11 Jul 2008 11:14:21 -0400 From: Stephen Hoffman Subject: Re: Changes to the freeware web site Message-ID: AEF wrote: > On Jul 10, 11:59 am, AEF wrote: >> On Jul 10, 9:34 am, "Tom Linden" wrote: >> >>> On Thu, 10 Jul 2008 02:18:37 -0700, Keith Cayemberg >>> wrote: >>>> Look for the following directories... >>>> ftp://mvb.saic.com/freewarev40/ >>>> ftp://mvb.saic.com/freewarev50/ >>>> ftp://mvb.saic.com/freewarev60/ >>>> ftp://mvb.saic.com/freewarev70/ >>>> ftp://mvb.saic.com/freewarev80/ I've been pointing folks over to Mark Berryman's Freeware site for a while now, given what's been going on with the re-organization and shuffling over at the HP web and HP ftp sites. Mark's ftp site tends to have better continuity, and has rather more content. >>> are these disjoint sets? JF, dude, you've been posting to comp.os.vms for, what, over a decade now? If you don't know how to answer this question for yourself by the expedient of looking at the directory and thinking about it for a couple of milliseconds, then there's not much help that can be provided. Caution: If you find this particular condition persists and wasn't just a case of the transient brain cramps that we all sometimes suffer from, we're going to be forced to assume you're eligible for senior management or even government office. Though unfortunately for this career change, you might already know too much about the internerds and the innertubes and these computer-thingies to reach your full ministerial- or C-level managerial potential. :-) If persistent eligibility for C-level or high office arises, please see your medical provider. :-) But rather more seriously, launch Finder and Cmd-K to connect up to the HP or to Mark's ftp server. You can then browse the site using Finder, and can drag-and-drop as needed. (I'm assuming you're not on a stone-age version of Mac OS here.) But then, you probably knew this... >>... I believe they are. In order to keep it down to two CD's (or whatever) >> Hoff decided to only include updates and new packages. That's probably >> why you don't see 1 thru 3. So if you're looking for the latest >> version of a package, start with 8 and work your way back. >> >> AEF > > > Verified! > > See > > http://groups.google.com/group/comp.os.vms/msg/fd919f82d1e8992e > > And I assume this holds for V6 and up, also. V8 involved substantial submission scrounging; there were far, far, far fewer submissions than had arrived for previous distributions; the rate completely and unexpectedly cratered. Accordingly, V8 has incorporated the contents of a few of the Freeware web sites (I'd asked Hunter about and subsequently replicated the contents of the Process Freeware site, for instance). There some repeats. But then the V8 distro still didn't start with V7; it was disjoint. Here's V3: I'd asked one of the HP folks about Freeware distribution plans recently, given I might have some Freeware to submit by then. There are apparently no plans as yet, so (depending on my schedule for these bits) I may end up tossing the bits over to Hunter for his Process ftp site. -- www.HoffmanLabs.com Services for OpenVMS ------------------------------ Date: Fri, 11 Jul 2008 08:38:57 -0700 From: "Tom Linden" Subject: Re: Changes to the freeware web site Message-ID: On Fri, 11 Jul 2008 08:14:21 -0700, Stephen Hoffman wrote: > JF, dude, you've been posting to comp.os.vms for, what, over a decade > now? It wasn't JF, it was lazy Tom. -- PL/I for OpenVMS www.kednos.com ------------------------------ Date: Fri, 11 Jul 2008 11:50:59 -0400 From: Stephen Hoffman Subject: Re: Changes to the freeware web site Message-ID: Tom Linden wrote: > On Fri, 11 Jul 2008 08:14:21 -0700, Stephen Hoffman > wrote: > >> JF, dude, you've been posting to comp.os.vms for, what, over a decade >> now? > > It wasn't JF, it was lazy Tom. Ok. Whoops. Sorry, JF. Maybe I'm eligible for high office, too. (A distant cousin of mine isn't the sharpest tool in the shed, and he's gotten himself elected.) -- www.HoffmanLabs.com Services for OpenVMS ------------------------------ Date: Fri, 11 Jul 2008 08:36:57 -0700 From: "Tom Linden" Subject: DCL divide by zero WAS: Symbol Substitution Mystery Message-ID: On Fri, 11 Jul 2008 08:09:09 -0700, AEF wrote: > I would definitely want to know as you can get very wrong results, but > I'd also want to know about overflow, which also isn't checked for, > and is much harder for the user to detect. > [...] Maybe you should be using a programming language, since DCL clearly was not intended for such algorithms -- PL/I for OpenVMS www.kednos.com ------------------------------ Date: Fri, 11 Jul 2008 11:33:21 -0400 From: Stephen Hoffman Subject: Re: DECW$SERVER_0 at 100% CPU? Message-ID: Christoph Gartmann wrote: > In article <4875e87c$0$18525$c3e8da3@news.astraweb.com>, JF Mezei writes: >> show process/cont decw$server_0 and look at the page file quota. If >> running mozilla on VMS, you need to give both the decw$server and the >> mosilla process gigantic page file quotas because of memory leaks and >> the fact that mozilla "forgets" to destroy Xpixmaps when you leave a >> page, and those remain in the decw$server's memory. > > Uh, I see. Any chance that there is a newer version of CWSB in the near future? Unlikely, given Mozilla walked away from the project. That written, the Seamonkey project (recently releasing its 1.1.10) is arguably the direct descendant of the old Seahorse ("Mozilla Suite") project upon which the old Compaq Secure Web Browser (CSWB, SWB) is based. The OpenVMS roadmap when last I checked it pointed to Mozilla Firefox and Thunderbird as a planned and potential future option. These days, building a browser UI for the WebKit engine would seen as (or more?) feasible. Some of what I've seen within the SWB source tree is a little bit gnarly. -- www.HoffmanLabs.com Services for OpenVMS ------------------------------ Date: 11 Jul 2008 06:49:45 -0500 From: clubley@remove_me.eisner.decus.org-Earth.UFP (Simon Clubley) Subject: Re: MTH$ RTL documentation Message-ID: <+hI7po+VUHvE@eisner.encompasserve.org> In article , "Richard Maher" writes: > Hi, > > Can someone please tell me where the online documentation for the MTH$ RTL > can be found? > > It's not in the "obsolete" section with PPL$ and DecTalk, but then I can't > find it in the "normal" docs either :-( > > Looking for mth$random at the mo, but am also interested in all the lovely > mth$jiand etc routines. > > Cheers Richard maher > It's right there on the website, but under the V7.3 section: http://h71000.www7.hp.com/doc/73final/6117/6117PRO.html The routines that you mention above are listed in it. Simon. -- Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP Microsoft: Bringing you 1980's technology to a 21st century world ------------------------------ Date: 11 Jul 2008 07:01:14 -0500 From: clubley@remove_me.eisner.decus.org-Earth.UFP (Simon Clubley) Subject: Re: MTH$ RTL documentation Message-ID: In article <+hI7po+VUHvE@eisner.encompasserve.org>, clubley@remove_me.eisner.decus.org-Earth.UFP (Simon Clubley) writes: > > It's right there on the website, but under the V7.3 section: > > http://h71000.www7.hp.com/doc/73final/6117/6117PRO.html > > The routines that you mention above are listed in it. > Never mind; I see that you already got an answer in another thread. However, for future reference, Google will locate the manual for you: Try entering "site:h71000.www7.hp.com/doc mth$random" (without the quotes) into Google. For me, the relevant manual is on the first page. Simon. -- Simon Clubley, clubley@remove_me.eisner.decus.org-Earth.UFP Microsoft: Bringing you 1980's technology to a 21st century world ------------------------------ Date: Fri, 11 Jul 2008 18:32:16 +0800 From: "Richard Maher" Subject: Re: Quasi-push technologies Message-ID: Hi Jeffrey, > Originally the ticker on the Monex site was an applet, but it got > replaced by flash when the site was overhauled. Sorry, poor quoting on my part. I was referring to the original: - >> See http://www.sgx.com/ > The only "push" > technology I know of is either Comet (Ajax with a persistent connection) Yes, the sow's ear (only with new clothes!) That is the sort of state-of-the-art http bollocks that I expected they'll all be gaggin' for :-( > or some form of Java (that again has a persistent connection). Otherwise > you have to have something timed that periodically checks for updates. Ah, but there is a third way; more soon. Cheers Richard Maher PS. Looks like Java's made it to the iPhone after all? "Jeffrey H. Coffield" wrote in message news:zDddk.13744$jI5.5318@flpi148.ffdc.sbc.com... > > > Richard Maher wrote: > > Hi Jeffrey, > > > > Thanks for the reply. > > > >> I can't easily tell if this is the same concept, but look at > >> www.monex.com. That ticker simply reads a file every few minutes that is > >> updated by the trading system which is an OpenVMS cluster. > > > > Interesting; you wouldn't happen to know if it's a periodic refresh, or Ajax > > or a hidden iFrame would you? (Or something else?) > > > > It's just that I have the Java console turned on for Applets and it popped > > up on that page, and I couldn't see an obvious Applet Object def (that > > wasn't buried in son included .JS file) and was just curious as to which bit > > the Java was controlling. > > > > Cheers Richard Maher > > > > PS. I'm currently very interested in various "push" technologies. (Used > > mainly in conjunction with Tier3 client/server request/response mechanisms) > > > > "Jeffrey H. Coffield" wrote in message > > news:nJXck.13044$jI5.11126@flpi148.ffdc.sbc.com... > >> > >> Richard Maher wrote: > >>> Hi, > >>> > >>> "Jan-Erik Söderholm" wrote in message > >>> news:5CHck.777$U5.393@newsb.telia.net... > >>>> See http://www.sgx.com/ > >>>> Click "SGX moves to new securities trading engine". > >>>> > >>>> Or see : http://tinyurl.com/5zpm8k for a > >>>> less-capable-browser-version... > >>> Great news! Especially as this is after the Nasdaq OMX acquisition. > >>> > >>> Cheers Richard Maher > >>> > >>> PS. If anyone knows (or bothers to work out) what the Java Applet is for > > on > >>> that page then please let me know. (Live ticker "push" technology?) > >>> > >>> > >> I can't easily tell if this is the same concept, but look at > >> www.monex.com. That ticker simply reads a file every few minutes that is > >> updated by the trading system which is an OpenVMS cluster. > >> > >> Jeff Coffield > >> www.digitalsynergyinc.com > > > > > Originally the ticker on the Monex site was an applet, but it got > replaced by flash when the site was overhauled. The only "push" > technology I know of is either Comet (Ajax with a persistent connection) > or some form of Java (that again has a persistent connection). Otherwise > you have to have something timed that periodically checks for updates. > > Jeff C. ------------------------------ Date: Fri, 11 Jul 2008 18:51:17 +0800 From: "Richard Maher" Subject: Re: Quasi-push technologies Message-ID: Hi (again) Jefferey, > We (Digital Synergy) have developed a Java to RMS connector that I am > working toward releasing open source. Sounds interesting. I look forward to seeing it. > it is capable of having the server push content to the client app. It > also allows Java clients to call OpenVMS legacy code. If you are > interested, look at http://www.digitalsynergyinc.com/JavaRMS.html. As I've said, much more on the push-side of things soon; but some of you from planet koozbane may wish to also consult: - http://manson.vistech.net/t3$examples/ In particular: - http://manson.vistech.net/t3$examples/demo_client_flex.html and http://manson.vistech.net/t3$examples/demo_client_web.html As you can see, all source code available, and PDF documentation included. Username: TIER3_DEMO Password: QUEUE Apparently, the cognoscente at HP/VMS have had much better solutions (and have had for ten years!) but it's just that they're not allowed to tell anyone about them. (Either that or they're a bunch of useless wankers that get paid for doing fuck-all - you decide!) Cheers Richard Maher "Jeffrey H. Coffield" wrote in message news:sJddk.11901$cW3.7340@nlpi064.nbdc.sbc.com... > > > Richard Maher wrote: > > > PS. I'm currently very interested in various "push" technologies. (Used > > mainly in conjunction with Tier3 client/server request/response mechanisms) > > We (Digital Synergy) have developed a Java to RMS connector that I am > working toward releasing open source. Since it doesn't use SQL or JDBC, > it is capable of having the server push content to the client app. It > also allows Java clients to call OpenVMS legacy code. If you are > interested, look at http://www.digitalsynergyinc.com/JavaRMS.html. > > Jeff Coffield ------------------------------ Date: Fri, 11 Jul 2008 01:00:47 -0700 (PDT) From: AEF Subject: Re: Symbol Substitution Mystery Message-ID: <69545e9c-e3f3-4da7-94d8-5098b23397d8@d1g2000hsg.googlegroups.com> On Jul 10, 10:19 pm, Rob Brown wrote: > On Thu, 10 Jul 2008, AEF wrote: > > On Jul 10, 2:02 pm, Rob Brown wrote: > >> On Wed, 9 Jul 2008, AEF wrote: > >>> Yes, that makes sense -- especially if division is performed by > >>> repeatedly subtracting .... (If that's not how it's done I'm all > >>> ears.) > > >> I might have thought that it would use the divide instruction. On > >> VAX the integer DIV instruction does one of two specific things on > >> divide by 0, neither of which is to return 2**31 - 1. On Alpha, > >> there is no > > > So what are these two specific things? > > The quotient operand is either unchanged or gets the value from the > divident operand. Which really means that it always gets the > dividend. I don't understand this. You're starting with the answer, no? > > >> integer divide instruction, but floating divide by zero causes a > > > You mean you can't run the command $ A = 5/2 on an Alpha? Surely you > > jest. > > Sure you can. Watch: > > $ write sys$output f$getsyi ("hw_name") > DEC 3000 Model 400 > $ a = 5/2 > $ show symbol a > A = 2 Hex = 00000002 Octal = 00000000002 > > Which does not change the fact that this machine does not have an > integer divide instruction. How can you perform integer division without being able to do it? Perhaps I'm misunderstanding just what you mean by "instruction". A computer can't do anything without instructions, no? > > >> "Division by Zero Arithmetic Trap" and the value in the result > >> register is UNPREDICTABLE. > > > You mean it spits out garbage? > > Garbage in, garbage out. Well, this means you have to do your own special-casing. I suppose it's not any worse than having to deal with DCL not giving you any warning when your operation overflows. > > I don't think "spits out" is accurate. The value is in the result > register, where you asked it to be. Well, you didn't say just what *does* happen to it. > > >>> I would have hoped that division by zero would have been > >>> special-cased. > > >> Looks to me like returning 2**31 - 1 is special-casing divide by 0. > > > I meant special-cased to not give a ridiculous answer. > > What answer to divide by 0 is less ridiculous than 2**31 - 1? After > all, the limit of a/x as x approaches 0 is infinity. You are not > going to get any closer to infinity than 2**31 - 1 with a 32 bit > signed integer. Only if a and x are of the same sign. If they are of different sign, the limit is -oo. And you can't get any farther from -oo than 2**31-1 with a 32-bit signed integer. And if a=0, none of this, including your statement, applies, because in that case the limit is zero, which is only one of an inifinite number of solutions. [Sometimes you have to be very careful with limits. There is the interesting example in Thomas (Calculus) wherein a cylinder is inscribed by a particular pattern of triangles with parameters m and n. If you take the limit to oo of one of the parameters first (say m for definiteness), then you get infinity instead of the right answer. This is because in that case the inscribed surface (composed of the triangles) does not approach the cylinder (which is certainly not obvious without some careful thought). But if you take the limit to oo of n first, or of m and n in a fixed ratio, you get the right answer. A very interesting example!] Regardless, any number as a result for division by zero is ridiculous. For a/0 with a.NE.0 there is no solution and there is no unique limit. For 0/0 there are an infinite number of solutions. The only non- ridiculous possibilities I'm aware of are a fatal error and this strange NAN that I'm not very familiar with. > > >> In a perfect world, what would you like it to do? > > > $ A = 4/0 > > %DCL-F-DIVBYZERO, attempt to divide by zero > > Yeah. Too bad they didn't do that. > > -- > > Rob Brown b r o w n a t g m c l d o t c o m > G. Michaels Consulting Ltd. (780)438-9343 (voice) > Edmonton (780)437-3367 (FAX) > http://gmcl.com/ AEF ------------------------------ Date: Fri, 11 Jul 2008 05:10:41 -0700 (PDT) From: AEF Subject: Re: Symbol Substitution Mystery Message-ID: <70902b76-fcee-4fd7-ae53-180fe49d4609@z72g2000hsb.googlegroups.com> On Jul 11, 6:30 am, VAXman- @SendSpamHere.ORG wrote: > In article <69545e9c-e3f3-4da7-94d8-5098b2339...@d1g2000hsg.googlegroups.com>, AEF writes: > > >{...snip...} > > >> What answer to divide by 0 is less ridiculous than 2**31 - 1? After > >> all, the limit of a/x as x approaches 0 is infinity. You are not > >> going to get any closer to infinity than 2**31 - 1 with a 32 bit > >> signed integer. > > >Only if a and x are of the same sign. If they are of different sign, > >the limit is -oo. And you can't get any farther from -oo than 2**31-1 > >with a 32-bit signed integer. And if a=0, none of this, including your > >statement, applies, because in that case the limit is zero, which is > >only one of an inifinite number of solutions. > > ...and DCL attempts this too: > > $ WRITE SYS$OUTPUT 1/0 > 2147483647 > $ WRITE SYS$OUTPUT -1/0 > -2147483648 OK, so DCL gives either of two ridiculous answers. But even here, assuming the form a/x, where a.LT.0, you still don't have a unique limit because x can approach a through negative values which would give +oo. The algorithm I suggested would produce the same answers that we see here that DCL does. > > -- > VAXman- A Bored Certified VMS Kernel Mode Hacker VAXman(at)TMESIS(dot)COM [...] AEF ------------------------------ Date: Fri, 11 Jul 2008 05:13:50 -0700 (PDT) From: AEF Subject: Re: Symbol Substitution Mystery Message-ID: <0e9e92ba-f9b4-4317-a193-bb41bd85a74b@j22g2000hsf.googlegroups.com> On Jul 11, 7:59 am, "Richard B. Gilbert" wrote: > AEF wrote: > > On Jul 10, 10:19 pm, Rob Brown wrote: > >> On Thu, 10 Jul 2008, AEF wrote: > >>> On Jul 10, 2:02 pm, Rob Brown wrote: > >>>> On Wed, 9 Jul 2008, AEF wrote: > >>>>> Yes, that makes sense -- especially if division is performed by > >>>>> repeatedly subtracting .... (If that's not how it's done I'm all > >>>>> ears.) > >>>> I might have thought that it would use the divide instruction. On > >>>> VAX the integer DIV instruction does one of two specific things on > >>>> divide by 0, neither of which is to return 2**31 - 1. On Alpha, > >>>> there is no > >>> So what are these two specific things? > >> The quotient operand is either unchanged or gets the value from the > >> divident operand. Which really means that it always gets the > >> dividend. > > > I don't understand this. You're starting with the answer, no? > > >>>> integer divide instruction, but floating divide by zero causes a > >>> You mean you can't run the command $ A = 5/2 on an Alpha? Surely you > >>> jest. > >> Sure you can. Watch: > > >> $ write sys$output f$getsyi ("hw_name") > >> DEC 3000 Model 400 > >> $ a = 5/2 > >> $ show symbol a > >> A = 2 Hex = 00000002 Octal = 00000000002 > > >> Which does not change the fact that this machine does not have an > >> integer divide instruction. > > > How can you perform integer division without being able to do it? > > Perhaps I'm misunderstanding just what you mean by "instruction". A > > computer can't do anything without instructions, no? > > You can program around a missing divide instruction. Iterated > subtraction will do it if you can't think of any other way. Many of the > early computers did not have a divide instruction and the pioneers had > to "roll their own". I've forgotten the details but, if anyone really > cares, a search through something like "Fundamental Algorithms" and/or > "Semi-Numerical Algorithms" by Donald E. Knuth should prove fruitful. OK, so I guess that by "instruction" it is meant that the processor itself has its own algorithm. And in the case where it doesn't, DCL itself has to use its own algorithm and issue addition and subtraction commands to the processor. My apologies for not being up on this lingo. OK. AEF ------------------------------ Date: Fri, 11 Jul 2008 10:06:21 -0400 From: norm.raphael@metso.com Subject: Re: Symbol Substitution Mystery Message-ID: This is a multipart message in MIME format. --=_alternative 004DB54685257483_= Content-Type: text/plain; charset="US-ASCII" AEF wrote on 07/11/2008 08:13:50 AM: > On Jul 11, 7:59 am, "Richard B. Gilbert" > wrote: > > AEF wrote: > > > On Jul 10, 10:19 pm, Rob Brown wrote: > > >> On Thu, 10 Jul 2008, AEF wrote: > > >>> On Jul 10, 2:02 pm, Rob Brown wrote: > > >>>> On Wed, 9 Jul 2008, AEF wrote: > > >>>>> Yes, that makes sense -- especially if division is performed by > > >>>>> repeatedly subtracting .... (If that's not how it's done I'm all > > >>>>> ears.) > > >>>> I might have thought that it would use the divide instruction. On > > >>>> VAX the integer DIV instruction does one of two specific things on > > >>>> divide by 0, neither of which is to return 2**31 - 1. On Alpha, > > >>>> there is no > > >>> So what are these two specific things? > > >> The quotient operand is either unchanged or gets the value from the > > >> divident operand. Which really means that it always gets the > > >> dividend. > > > > > I don't understand this. You're starting with the answer, no? > > > > >>>> integer divide instruction, but floating divide by zero causes a > > >>> You mean you can't run the command $ A = 5/2 on an Alpha? Surely you > > >>> jest. > > >> Sure you can. Watch: > > > > >> $ write sys$output f$getsyi ("hw_name") > > >> DEC 3000 Model 400 > > >> $ a = 5/2 > > >> $ show symbol a > > >> A = 2 Hex = 00000002 Octal = 00000000002 > > > > >> Which does not change the fact that this machine does not have an > > >> integer divide instruction. > > > > > How can you perform integer division without being able to do it? > > > Perhaps I'm misunderstanding just what you mean by "instruction". A > > > computer can't do anything without instructions, no? > > > > You can program around a missing divide instruction. Iterated > > subtraction will do it if you can't think of any other way. Many of the > > early computers did not have a divide instruction and the pioneers had > > to "roll their own". I've forgotten the details but, if anyone really > > cares, a search through something like "Fundamental Algorithms" and/or > > "Semi-Numerical Algorithms" by Donald E. Knuth should prove fruitful. > > OK, so I guess that by "instruction" it is meant that the processor > itself has its own algorithm. And in the case where it doesn't, DCL > itself has to use its own algorithm and issue addition and subtraction > commands to the processor. My apologies for not being up on this > lingo. Well, we're more than a little off-topic here, but... In my early employment I worked on an IBM 1620 machine without a hardware divide instruction. That instruction was emulated in software in the assembler. A later model implemented it in hardware. (We also used FORTRAN I, which did all floating-point with subroutines.) The console was and IBM electric typewriter, and the output was a card-punch, There was no printer; we used an IBM 407 to list the card deck. Others have recalled here those ancient devices. There is nothing unusual about this. Divide-by-zero has been a thorn in the side of programmers since there were programmers. I don't think I want DCL to stop on a runtime error like that, but maybe it should generate an informational warning, since I'd probably want to know my calculations were suspect at best. > > OK. > > AEF --=_alternative 004DB54685257483_= Content-Type: text/html; charset="US-ASCII"
AEF <spamsink2001@yahoo.com> wrote on 07/11/2008 08:13:50 AM:

> On Jul 11, 7:59 am, "Richard B. Gilbert" <rgilber...@comcast.net>
> wrote:
> > AEF wrote:
> > > On Jul 10, 10:19 pm, Rob Brown <mylastn...@gmcl.com> wrote:
> > >> On Thu, 10 Jul 2008, AEF wrote:
> > >>> On Jul 10, 2:02 pm, Rob Brown <mylastn...@gmcl.com> wrote:
> > >>>> On Wed, 9 Jul 2008, AEF wrote:
> > >>>>> Yes, that makes sense -- especially if division is performed by
> > >>>>> repeatedly subtracting .... (If that's not how it's done I'm all
> > >>>>> ears.)
> > >>>> I might have thought that it would use the divide instruction.  On
> > >>>> VAX the integer DIV instruction does one of two specific things on
> > >>>> divide by 0, neither of which is to return 2**31 - 1.  On Alpha,
> > >>>> there is no
> > >>> So what are these two specific things?
> > >> The quotient operand is either unchanged or gets the value from the
> > >> divident operand.  Which really means that it always gets the
> > >> dividend.
> >
> > > I don't understand this. You're starting with the answer, no?
> >
> > >>>> integer divide instruction, but floating divide by zero causes a
> > >>> You mean you can't run the command $ A = 5/2 on an Alpha? Surely you
> > >>> jest.
> > >> Sure you can.  Watch:
> >
> > >>    $ write sys$output f$getsyi ("hw_name")
> > >>    DEC 3000 Model 400
> > >>    $ a = 5/2
> > >>    $ show symbol a
> > >>      A = 2   Hex = 00000002  Octal = 00000000002
> >
> > >> Which does not change the fact that this machine does not have an
> > >> integer divide instruction.
> >
> > > How can you perform integer division without being able to do it?
> > > Perhaps I'm misunderstanding just what you mean by "instruction". A
> > > computer can't do anything without instructions, no?
> >
> > You can program around a missing divide instruction.  Iterated
> > subtraction will do it if you can't think of any other way.  Many of the
> > early computers did not have a divide instruction and the pioneers had
> > to "roll their own".  I've forgotten the details but, if anyone really
> > cares, a search through something like "Fundamental Algorithms" and/or
> > "Semi-Numerical Algorithms" by Donald E. Knuth should prove fruitful.
>
> OK, so I guess that by "instruction" it is meant that the processor
> itself has its own algorithm. And in the case where it doesn't, DCL
> itself has to use its own algorithm and issue addition and subtraction
> commands to the processor. My apologies for not being up on this
> lingo.


Well, we're more than a little off-topic here, but...

In my early employment I worked on an IBM 1620 machine without a hardware
divide instruction.  That instruction was emulated in software in the
assembler.  A later model implemented it in hardware.  (We also used
FORTRAN I, which did all floating-point with subroutines.)  The console
was and IBM electric typewriter, and the output was a card-punch, There
was no printer; we used an IBM 407 to list the card deck.  Others have
recalled here those ancient devices.

There is nothing unusual about this.  Divide-by-zero has been a thorn in
the side of programmers since there were programmers.  I don't think I
want DCL to stop on a runtime error like that, but maybe it should
generate an informational warning, since I'd probably want to know my
calculations were suspect at best.
>
> OK.
>
> AEF
--=_alternative 004DB54685257483_=-- ------------------------------ Date: Fri, 11 Jul 2008 07:27:57 -0700 From: "Tom Linden" Subject: Re: Symbol Substitution Mystery Message-ID: On Fri, 11 Jul 2008 07:06:21 -0700, wrote: > There is nothing unusual about this. Divide-by-zero has been a thorn in > the side of programmers since there were programmers. I don't think I > want DCL to stop on a runtime error like that, but maybe it should > generate an informational warning, since I'd probably want to know my > calculations were suspect at best. doesn't ON ERROR or ON SEVERE_ERROR give you that functionality? -- PL/I for OpenVMS www.kednos.com ------------------------------ Date: Fri, 11 Jul 2008 08:09:09 -0700 (PDT) From: AEF Subject: Re: Symbol Substitution Mystery Message-ID: <7edfdb47-a90f-44e3-8031-07256c99d2a5@r66g2000hsg.googlegroups.com> On Jul 11, 10:06 am, norm.raph...@metso.com wrote: > AEF wrote on 07/11/2008 08:13:50 AM: > > > > > > > On Jul 11, 7:59 am, "Richard B. Gilbert" > > wrote: > > > AEF wrote: > > > > On Jul 10, 10:19 pm, Rob Brown wrote: > > > >> On Thu, 10 Jul 2008, AEF wrote: > > > >>> On Jul 10, 2:02 pm, Rob Brown wrote: > > > >>>> On Wed, 9 Jul 2008, AEF wrote: [...stuff about DCL producing actual numbers as answers for dividing by zero and doing so without signaling an error and stuff about divide instructions omitted...] > Well, we're more than a little off-topic here, but... > > In my early employment I worked on an IBM 1620 machine without a hardware > divide instruction. That instruction was emulated in software in the > assembler. A later model implemented it in hardware. (We also used > FORTRAN I, which did all floating-point with subroutines.) The console > was and IBM electric typewriter, and the output was a card-punch, There > was no printer; we used an IBM 407 to list the card deck. Others have > recalled here those ancient devices. > > There is nothing unusual about this. Divide-by-zero has been a thorn in > the side of programmers since there were programmers. I don't think I > want DCL to stop on a runtime error like that, but maybe it should > generate an informational warning, since I'd probably want to know my > calculations were suspect at best. I would definitely want to know as you can get very wrong results, but I'd also want to know about overflow, which also isn't checked for, and is much harder for the user to detect. [...] AEF ------------------------------ Date: Fri, 11 Jul 2008 11:44:56 -0400 From: Stephen Hoffman Subject: Re: Symbol Substitution Mystery Message-ID: Tom Linden wrote: > On Fri, 11 Jul 2008 07:06:21 -0700, wrote: > >> There is nothing unusual about this. Divide-by-zero has been a thorn in >> the side of programmers since there were programmers. I don't think I >> want DCL to stop on a runtime error like that, but maybe it should >> generate an informational warning, since I'd probably want to know my >> calculations were suspect at best. > > doesn't ON ERROR or ON SEVERE_ERROR give you that functionality? Nope. Not on OpenVMS Alpha V7.3-2. The ON WARNING (the lowest you can go) doesn't fire for a constant DBZ. (It is also feasible to test this DCL on the OpenVMS version that most interests you, to see if this DCL behavior has changed.) If you're doing this sort of thing for real, you might want to look at another of the available scripting languages. I haven't tried bash here, but that or Python or Perl are obvious potential options. The OpenVMS Ruby port is somewhat stale, so that's probably not the best choice. -- www.HoffmanLabs.com Services for OpenVMS ------------------------------ End of INFO-VAX 2008.385 ************************