Subject: Re: RFC: recommended dependencies (diffs attached)
To: Rene Hexel <firstname.lastname@example.org>
From: Jeremy C. Reed <email@example.com>
Date: 01/07/2004 23:42:41
On Thu, 8 Jan 2004, Rene Hexel wrote:
> Current practice is to bump PKGREVISIONs and
> dependencies for all directly or indirectly dependent
> packages, regardless of whether there actually is any
> technical requirement to bump dependencies or not.
I don't think that is the current practice. If that is done that is wrong
and we should make sure that doesn't happen in the first place.
For tiff, the technical requirement was that on some platform(s), it
did not work.
> While this helps binary package consistency, it also
> causes a lot of overhead with users/developers who compile
> packages from source. Often huge portions of an installed
> package base need to be recompiled unnecessarily.
I agree. It is an inconvenience and I have complained too.
> I have attached diffs to pkgsrc that do the following:
> - Introduce a new variable, RECOMMENDED, that has
> the same syntax as DEPENDS.
RECOMMENDED (without looking at your patches) reminds me of Jan
Schaumann's RFC of variable "SUGGESTS".
> If someone doesn't want to (or can't afford to)
> update tiff and a huge number of packages straight
> away, they can set IGNORE_RECOMMENDED=yes and still
> compile perfectly working binaries against their
> installed package base. But they will be warned
> that their binaries may be incompatible with
> binaries compiled against the latest version of
> tiff and should therefore not be uploaded.
That sounds interesting.
Also, what about actually automatically registering libraries so names
(and/or other files needed)? (I mentioned that in an email, maybe last
Jeremy C. Reed