On Tue, Jun 09, 2020 at 08:21:12PM +0200, Marc Espie wrote: > On Tue, Jun 09, 2020 at 07:52:53PM +0200, Rafael Sadowski wrote: > > On Tue Jun 09, 2020 at 11:35:12AM -0600, Antoine Jacoutot wrote: > > > CVSROOT: /cvs > > > Module name: ports > > > Changes by: [email protected] 2020/06/09 11:35:12 > > > > > > Modified files: > > > x11/qt5/qtwebkit: Makefile > > > > > > Log message: > > > Unbreak PKGNAME. > > > > What was the problem? The EPOCH bump was intended and well tested. > > No, you can't do that. > > The only mechanism to do EPOCH bumps is through... EPOCH > > Rolling your own by embedding v0 in the PKGNAME is entirely unsupported > and *will* break various things. > > Specifically, revision gets added in the wrong location, and pkg_create(1) > rightfully protests, which is cool, because it would be much more annoying > to fix later. > > Note that at the moment, I don't quite remember all the reasons why rolling your own is bad, but I made a very conscious decision to architect all the PKGNAME/FULLPKGNAME/REVISION/EPOCH/MULTI_PACKAGES mess the way it is designed, and it is 100% unsustainable and not supported to try to do things the way you do.
If anything, I would add more checks to prevent the things you try from accidentally working if I could. If you want an EPOCH, you use EPOCH. If that interferes with dependencies PKGSPEC was written precisely to handle that issue.
