Re: CONFLICTS variable is terrible

[Aleksey Cheusov <cheusov%tut.by@localhost>]

> On Mon, Nov 03, 2008 at 07:29:42PM +0200, Aleksey Cheusov wrote:
 >>  >> 1) TONS of packages do not set CONFLICTS variable properly.
 >> 
 >> > Yes, it is.
 >> I personally think that missed CONFLICTS is critical bug and
 >> most of them MUST be fixed before 2008Q4 happens.

> The criticality (-ness? -hood?) of bug is, thankfully, subjective.
Yes, but...
Suppose you've installed lang/js and then lang/ossp-js.
What will be in PREFIX/bin/js?
Now suppose you use higher level of package management tool,
say, pkg_chk. What will be in PREFIX/bin/js now? Random file.
I treat this bug critical.

 >> What's your plans?
 >> For this one day noone even tried to use/test wip/pkg_conflicts.

> Sending a 1000-odd line post to the lists, several times, doesn't
> exactly sound like good publicity to me.
Do you see a better way? I've found a bug which is about 1000 lines in
size.  How to let you know about it?

 >> > How to detect CONFLISTS before installation from source?
 >> Filenames from currently installed package (before unpacking)
 >> should be checked against filenames stored in database.

> I honestly don't know how to qualify that proposition in a way that
> doesn't sound offensive.  So I'll just say it's completely stupid.
I knew it :-)

> Package conflicts and dependencies are the same information, they're
> needed at the metadata level for any basic package management.
As I altready said, there is another way.
In my view, in a long run, it is better.

> Whether or not we can obtain information about conflicts in a
> somewhat automated way is a different issue.
Yes, and provided a tool for this. It is good enough
at least it is for beginning.

> But the current
> situation certainly isn't as scandalous as you seem to believe.
> Now, if you manage to get some people interested in actually solving
> the issues you expose instead of just stating they exist (in a
> rather obnoxious way, I have to say), that would be a step forward.
Or my god. You are probably right (as always?).  Everything I did for
NetBSD and pkgsrc (packages, patches, tons of bug reports, bulk builds
under platform etc.) I did just to blame NetBSD and pkgsrc developers
and your personally. This is your constructive idea. Right?

> There is definitely a lot of information we could get from bulk
> build results that we currently don't retrieve, and a fortiori don't
> use.  (Although it must still be taken with care; a bulk build is
> specific to the platform it has run on.)
Lots of information can be retrieved with a help of wip/pkg_summary-utils.
Bulk builds are different question. See the logs of my distbb, for example.

-- 
Best regards, Aleksey Cheusov.