Re: Strange CONFLICTS definitions render pkgin full-upgrade impossible

[Jonathan Perkin <jperkin%joyent.com@localhost>]

* On 2021-05-21 at 12:31 BST, Aleksej Lebedev wrote:

> Your point is clear. Just one last though. It is a bug. I wanted to fix it but I couldn't understand the root cause until I looked at the source code of pkgin - only then I realized the problem is in the package itself.
> A better message from pkgin would save a few hours of my time.

So there are two better places to do work here if we want to avoid this 
problem in the future:

 - For each proposed CONFLICTS entries, perform a pmatch against the
   target PKGNAME during pkg_create, and abort if there is a match.
   This will ensure we never create and distribute a broken package.

 - Add a similar check to pkglint based on Makefile contents.  This will
   ensure we never commit a broken package, assuming the committer runs
   pkglint before committing...

The pkg_create patch should be trivial.  The pkglint patch may be more 
complex and may not catch all cases, as it would need to compute all 
possible PKGNAME variations.

--
Jonathan Perkin  -  Joyent, Inc.  -  www.joyent.com