Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!usc!elroy.jpl.nasa.gov!decwrl!shelby!portia!dhinds From: dhinds@portia.Stanford.EDU (David Hinds) Newsgroups: comp.sys.ibm.pc Subject: Re: programs crashing in Desqview Summary: Such is life... Message-ID: <9999@portia.Stanford.EDU> Date: 9 Mar 90 07:51:42 GMT References: <4913c88c.20b6d@apollo.HP.COM> Sender: David Hinds Distribution: usa Organization: Stanford University Lines: 56 In article <4913c88c.20b6d@apollo.HP.COM>, nelson_p@apollo.HP.COM (Peter Nelson) writes: > > As a result of all this I am forced to sadly conclude that > Quarterdeck's product, while probably adequate for certain > "known-to-be-safe" applications, is not yet a reasonable > choice for a general-purpose environment. Given Desqview's enormous installed base, is it any surprise that "many" people have had problems with it, given the extreme heterogeneity of DOS programs and environments? I consider it to be amazingly good at doing what seems at first to be a truly impossible task. Think about trying to write an operating system-level product to insulate a vast and unknown set of programs from doing bad things to a vast and unknown set of possible hardware configurations, when the programs expect to be able to do as they please. > > I was relating my Desqview woes to a friend at Lotus who said > he's heard it before about Desqview and suggested I try a > product called VM-386. Does anybody know anything about > this product? I understand it doesn't do windows: I can live > without that; what I really want is to have multiple (2 is > probably sufficient) tasks running at the same time without > crashing into each other or doing anything else they wouldn't > do all by themselves. The main thing I want is NO hassles and > NO farting around tweaking things trying to get them to work. > He said VM-386 fits the bill. Does it? Who makes it? I > still have 60 days to get my money back on the Desqview. > Despite Desqview's (I think unavoidable) weaknesses, I would be surprised to hear of another program that does a better job of doing these things. I haven't used VM-386, but it claims to be completely bulletproof. This level of protection is certainly possible on a 386 system operating in Virtual 86 mode. However, I think you will find that this degree of protection limits the sorts of programs and hardware that can be supported. There will be programs that fail under VM-386; it simply guarantees that they won't bring down the rest of the system. I'll bet that far fewer programs run under VM-386 than can be (admittedly) "tweaked" to run under Desqview. From what I remember about your Zortech C graphics program problem, the 386 was trapping a protection violation of unknown origin. So, the program was executing instructions that violate the privileges given to a Virtual 86 task. Desqview normally traps these violations and performs an appropriate action, if it can tell what the program was trying to do. For example, all DOS and BIOS calls, as software interrupts, cause protection violations in V86 mode, and Desqview (or QEMM, actually) has to temporarily take control before invoking the requested DOS service. If Desqview is virtualizing screen I/O, all accesses to display memory cause protection violations. Depending on how you set the "protection level" option, other sorts of actions can cause faults, including direct access to I/O ports via IN and OUT instructions, or even some innocuous-looking attempts to manipulate the flags register. I would be surprised if VM-386 could recover from the Zortech program faults; though they will not crash the system, I bet they won't run successfully. -David Hinds dhinds@popserver.stanford.edu