tech-pkg archive

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

Re: CONFLICTS variable is terrible

 >> I personally think that missed CONFLICTS is critical bug and
 >> most of them MUST be fixed before 2008Q4 happens.
 >> What's your plans?
 >> For this one day noone even tried to use/test wip/pkg_conflicts.

> I have no plan.  I don't konw why you want to fix it even though you
> want to remove the variable.

Because I don't believe this variable will be removed in nearest
future. Some of you think that DEPENDS are CONFLICTS are similar and
my proposal is stupidity.
I my view, CONFLICTS is completely different from DEPENDS.

Also "fixing" is not just adding new conflicts to the package.
Many packages should be fixed in a different way.
For example, lang/{js,spidermonkey,ossp-js}. See my previous email.
I think my original mail lists them all.

 >>  >> 2) I wrote wip/pkg_conflicts to automate checks for CONFLICTS.
 >>> It's nice, but seems some redundant.
 >>> For example, ap13-* v.s. ap22-*.
 >> I think wip/pkg_conflicts works correctly here.

 >> hint: In your case apache-1.3 and apache-2.2 do not conflict.

> apache-1.3 and apache-2.2 are same package with diffrent version,
> naturally conflict, no need to mark as CONFLICTS.
Ok. I already said wip/pkg_conflicts doesn't understand indirect
conflicts yet.

 >>  >> 3) Why not to remove CONFLICTS variable at all?
 >>> How to detect CONFLISTS before installation from source?
 >> Filenames from currently installed package (before unpacking) should be 
 >> checked
 >> against filenames stored in database.

> It's binary package behaviour.
> Again, "Install from source".
"Install from source" should be implemented as "Create binary package"
and then "Install binary". This is why DESTDIR appeared. And this is
pkgsrc's long-term goal. Right? Anyway do not create new unnecessary
entities. Package installation should be implemented in ONE way,
one way for binary package and 'make install'. Sooner or later.

 >>> And how about conflicts with older version, already resolved with
 >>> current version?
Yet another reason why CONFLICTS should be removed in the future.

Best regards, Aleksey Cheusov.

Home | Main Index | Thread Index | Old Index