[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Features.



On Wed, Aug 30, 2000 at 02:13:19AM +0200, Hubert Feyrer wrote:
> On Tue, 29 Aug 2000, Marc Espie wrote:
> > Right now, rollback is mostly necessary because the package is installed
> > directly without checking for problems first.
> 
> How do you want to check in advance for a user typing ^C in mid-pkg_add?

Obviously we can't. :)

However, I'm not really fond of the false sense of security rollback may
provide in such a case, imperfect as it is...

The thing is, the end user may somehow come to believe starting pkg_add and
typing ^C will accomplish complete rollback, and feel safe about it, and 
start depending on it...

One important feature of the pkg system is to try to make addition and removal
of components 100% safe. Each time we end up with an imcomplete system, where
the information in /var/db/pkg and what's actually installed doesn't match,
we have failed (at least that's the way I choose to see it).

> I guess I can claim that the NetBSD pkg_* have some tests in there, but I
> still feel safe to have the rollback functionality, see above. From the
> time when ^C handling we've learned that users *do* want to stop things.

It's quite possible... I still have trouble (and surprises) looking at 
what pkg_* claim to do versus what the code actually does... downright
unintuitive and convoluted in places.

Do you have a pointer to code that goes down a packing-list and checks 
that files will install without barfing before starting the install proper ?
Not saying it doesn't exist, but I haven't located it at a glance.
-- 
	Marc Espie		
|anime, sf, juggling, unicycle, acrobatics, comics...
|AmigaOS, OpenBSD, C++, perl, Icon, PostScript...
| `real programmers don't die, they just get out of beta'

To unsubscribe: send mail to <majordomo@unixathome.org>
with "unsubscribe bsdports" in the body of the message