One option smart needs to conquer the world

Pascal Bleser pascal.bleser at skynet.be
Wed Feb 22 14:58:59 PST 2006


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Gustavo Niemeyer wrote:
>> Alternative Policies for 'smart upgrade' would probably be a good
>> thing. For instance, I would really like a policy similar to 'apt
>> upgrade' that does not Remove packages. (apt dist-upgrade is used for
>> an upgrade that can install and remove packages)
>>
>> If there was an option that would prevent downgrading, then the
>> packager of smart for fedora could choose to enable it by default, if
>> they are so concerned. But users could choose what they want.

Keep in mind that there is always a good reason for removing or downgrading packages, it doesn't
happen just for fun (unless there would be a bug in smart's upgrade logic, which I doubt at this point).

The user is saying: I want to upgrade this and that.
Well, smart tries hard to do exactly that, even if it includes downgrading or removing other packages.

That sounds like exactly the correct behaviour.

Hence, the default behaviour should be like now.

What would be interesting though, especially for more experienced users actually, is to include the
possibility of trying to upgrade at best without downgrading and/or removing packages.

If it's not possible, then fail.

e.g. smart --no-downgrade upgrade kde* qt*
  or smart --no-downgrade --no-remove upgrade kde* qt*

If the engine cannot compute a path that allows upgrading the requested packages without removing
and/or downgrading, then it should fail.

But this should definitely *not* be the default behaviour. It's an option, at best.
Because what the user actually wants is to upgrade kde* and qt*

And of course, this should never be set globally as the default by the packager.
It's an end-user option.

> That's a crazy world. One of the reasons I wrote Smart was to avoid
> explosion when merely downgrading a package would fix the transaction.
> As a user (perhaps a non-conventional one), I always hated having to
> figure out why the transaction wouldn't work myself.

Yes. In some cases it could be interesting though, but as an option.
I don't consider this feature high priority.

I also sometimes see end-users "complaining" (kindof) on that, but once you explain to them how
dependencies work and why sometimes packages are removed or downgraded to satisfy dependencies, they
usually understand and don't "complain" any more.

> Is that what users want?  Not peforming an operation at all rather than
> being *ASKED* to downgrade a package?  If that's the case, I'll
> implement the flag in Smart that disables that support (and perhaps
> rename the project...).

Please don't :)
Smart is the name, because smart is the engine ;)

just my 0.02EUR

cheers
- --
  -o) Pascal Bleser     http://linux01.gwdg.de/~pbleser/
  /\\ <pascal.bleser at skynet.be>       <guru at unixtech.be>
 _\_v The more things change, the more they stay insane.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (GNU/Linux)

iD8DBQFD/Oyyr3NMWliFcXcRAi7mAJ9W/Z4j+Q4AvNLtdeSz2vZigJLViwCgj4fO
2kn0cUieGCoMn5hnAQqNXrk=
=Q2Z3
-----END PGP SIGNATURE-----



More information about the Smart mailing list