WARNING : Deinstall in SCO UNIX DANGEROUS!

[Wu Liu]

I wanted to let somebody else answer this one first.  Since nobody from
SCO Support (busy guys and gals that they are) has replied yet, though...

/--sef at kithrup.COM (Sean Eric Fagan) said...
| In article <9113 at lkbreth.foretune.co.jp> trebor at lkbreth.foretune.co.jp (Robert J Woodhead) writes:
| >Going into CUSTOM, I noticed that there were
| >several packages in the development system that I didn't need, like DOSDEV
| >and OS2DEV, so I decided to deinstall them.
| 
| I have deinstalled DOSDEV and OS2DEV on kithrup (several times, in fact, for
| reasons I won't get into here 8-)), and have never had any problems
| whatsoever.  I don't know what your problem was, but it's the first that I
| (even as an SCO employee) have heard of something like that happening.
\--

I've done multiple installations and deinstallations (I lost count long
ago...) of various SCO software products, at both the product and
package level, and I've never run into a deinstallation which trashed
the contents of /usr, like was reported.

/--
| >Upon trying to deinstall OS2DEV, CUSTOM reported an error.  
| 
| What error?  Did you, perhaps, run out of disk space?  (It tries to build a
| file list, according to its messages, and if it can only make a file that
| has "/usr" as the pathname...)
\--

Sean's on the right track here; the first step would be to determine
what happened to custom here.  There are two basic types of custom
errors.  The less severe type will print out some error message to the
screen, while the nastier ones will cause custom to exit prematurely.
The latter will also generate some nice, cryptic error messages like
"custom: Internal error 10"...

It would also help to know what versions of both the OS and Dev Sys
you have installed are.

/--
| >Things were trashed enough that I had to reinitialize the HD and reinstall
| >all the packages (sigh - an afternoon wasted).  
| 
| Ah.  I suspect from that that your disk actually *died*.  Not sco's fault.
| (Yes, such things happen occasionaly to disks.  One of the reinstalls for
| kithrup, a few months ago [and when I upgraded to unix] was because the disk
| died.  It reformated without problem, though, and a scan reported no bad
| tracks.)
| 
| >There were a few bobbles because
| >SCO doesn't bother to mention the dependencies -- who would imagine that to
| >run the C compiler you need the XENIX 386 cross development libraries?
| 
| You also have a *very* old version of the devsys.  You probably want to
| upgrade...
\--

I guess you've got version 3.2.0 of the Unix Dev Sys.  That's pretty
old stuff; I think the current shipping standalone version is 3.2.1,
maybe 3.2.2.  The version bundled with the Open Desktop Development
System currently is 3.2.1.

/--
| >Did anyone at SCO ever bother to test their install/deinstall scripts?
| 
| Gee, no.  Actually, we never install or deinstall internally, and you are,
| in fact, the first person to ever deinstall.  (That was sarcasm.)
\--

Now, Sean, lay off the poor customer.  After all, his money helps pay
your salary... :-)

As for the question itself; yes, we do bother to test the scripts.
However, it's virtually impossible to test all of the various
combinations of installations/deinstallations/mix of products.  We
also use our own products in-house, which catches quite a few bugs
that otherwise might have slipped past both engineering and QA tests.

Having said that, I feel pretty confident in saying that something
this obvious and this serious would have been caught long before the
product shipped.  Perhaps Mr. Woodhead could post or e-mail more
details about the problem he encountered?