To: mharris AT blackwidow DOT saultc DOT on DOT ca Cc: opendos AT delorie DOT com Subject: Re: Installation Problems Message-ID: <19970425.134647.6447.7.editor@juno.com> References: From: editor AT juno DOT com (Bruce Morgen) Date: Fri, 25 Apr 1997 13:48:25 EDT Precedence: bulk On Fri, 25 Apr 1997 11:22:19 -0400 (EDT) "Mike A. Harris" writes: > >A computer with 4DOS, is not an ordinary setup. It's common enough to be accounted for sensibly imho. >To use 4DOS as >your primary shell, you have to put it in a SHELL= line. Yup. >In >order for OD to function, it must change that line to point to >COMMAND.COM (for hopefully obvious reasons of a failsafe >install). It didn't do that Mike, it left the SHELL= line alone -- that's what caused the "corrupt or missing command interpreter" error message. If it had changed the SHELL= line I would have gotten a normal OD COMMAND.COM prompt and possibly some other errors on 4DOS-specific stuff in AUTOEXEC.BAT. >Since 4DOS (or any other shell, including even a >previous COMMAND.COM) is present on the SHELL line, it is renamed >to prevent problems with OD's COMMAND.COM. What possible "problems" could such an idiotic idea "prevent?" If the SHELL= line ends in a complete and valid pathname, the only problem I foresee involves secondary shells and the COMSPEC variable, not a previously used shell with an entirely different name and directory location. >This isn't the best, >or most innovative solution, but it does work. No it doesn't, it causes annoying problems if your current shell isn't COMMAND.COM. >If you understand >the hows and whys of the way the installer works it isn't hard to >see that it isn't random behaviour. Nobody said it was random, just badly designed and/or coded. >It works exactly the way >that they intended it to - albeit this is not very intuitively. > It works very badly, imho you are far too kind. Perhaps that is the way "they intended it to," but in that case their intentions weren't thought out very well at all. >> >I'm almost positive that a future version of OD will prevent this >> >sort of "brain damage" from happening. >> > >> Yes, if Caldera is aware of >> the installer's rude behavior >> I'm sure correcting it is >> something short of rocket >> science -- the way CONFIG.SYS >> is analyzed needs to be a tad >> more sophisticated, should be >> no big deal to fix. > >Or even simpler, put OD's COMMAND.COM into the directory that OD >is installed into, and then put SHELL=C:\OD\COMMAND.COM in there. Not as good, but at least there'd be an error-free boot-up. >This way it doesn't at ALL interfere with ANY alternate shells >and no coding effort is needed in the installer. I have allways >wondered why MSDOS, and other DOS's installed the command >interpreter in the root directory as well as in their respective >"\DOS" directories. It doesn't make any sense. I always remove >the root copy and point shell to the installed directory. You'd >think that most installations would be doing this as common >sense. > I agree, if SHELL= has a full and valid pathname, the extra copy of COMMAND.COM is superfluous, especially if the F5-on-bootup feature includes a default path of C:\DOS\. Perhaps the extra copy is to make sure there's there's a valid command interpreter in the event of a screw-up when using the F8-on-bootup line-by-line feature, where there's a chance of accidently skipping the SHELL= line in CONFIG.SYS. >> >Or, just rename it as I described above. >> > >> Sure, if you've figured out >> what and where the renamed >> and hidden file is. I've >> found no evidence of such >> on the XT, have yet to scan >> the 486. :-) > >Well, that is what it does if you care to look. Yes, it was an >annoyance, but the files are intact still nonetheless. > I looked, and either you're wrong or there is more than one installer glitch. >> >Or acquiring the COMMAND.COM source code and coding 4DOS features >> >into it yourself. I would suspect that by the time that OpenDOS >> >7.1 or 8.0 or whatever comes out that COMMAND.COM will not only >> >be compiled on a FREE compiler, but also may compile on ANY >> >compiler, and also will probably double or triple in size. >> >(executable size, not resident size). >> > >> Building up COMMAND.COM into >> something resembling a free >> 4DOS would be quite a project, >> but a reasonable 4DOS subset >> could be implemented without >> too much trouble imho -- sort >> of like the ZCPR series of >> command processor replacements >> for CP/M 2.2, which were no >> bigger than the original DRI >> CCP but much less annoying to >> use. > >Well, the source code for COMMAND.COM is now available, as are >the sources for various UNIX shells (bash, tcsh, pdksh, etc) so >the functionality is there, cloning 4DOS shouldn't be too >difficult to those with free time to spare. > Yup, free time is the main issue, most folks want to get paid for that much work, even if it's mostly cobbling together borrowed code. >> >No, I think the problem is no big deal, a lack of foresight on >> >the installer's part. No corruption. I even suspected >> >corruption at first, but when I figured it out, I ice cream coned >> >myself in the forehead. :o) >> > >> Right back atcha, Mike, >> thanks for detailing the >> renegade installer's >> alleged quirk(s). > >Yep, now we can put another couple requests on the wishlist. > >- Non braindamaged installer for OD. I think Caldera ought to do this. Right now the installer handles some things very well (e.g. it politely allows you to keep QEMM as you memory manager) and other things with notable incompetence (the one or more shell-related glitches). >- People willing to clone 4DOS into COMMAND.COM > I think people with lots of leisure time and/or little need for sleep ought to tackle this. >TTYL > bcnu -- Bruce