Subject: Re: Inconsistent dependency handling
To: Chapman Flack <email@example.com>
From: Gavan Fantom <firstname.lastname@example.org>
Date: 05/30/2006 09:43:49
Chapman Flack wrote:
> Gavan Fantom wrote:
>> Such a separation is not just philosophical. As soon as you have pkg_add
>> consulting the source, you lose the ability to bulk build on one
>> machine, and then install binary packages on many others without
>> installing the source. I would be very surprised if there aren't many
>> people out there doing just that right now.
> This is an argument of the form "if implemented badly, the idea wouldn't
> work," which is a species of strawman.
I understood the original mail to mean that pkg_add of a binary package,
in the general case, would require a pkgsrc tree around. That would be Bad.
> The use case where the current system loses badly is the case where
> the user does have pkgsrc installed and needs the system to efficiently
> install from binary where possible and build from source where
> necessary. The case is adequately covered by conditional behavior in
> pkg_add that is activated by a special option only when pkg_add is
> invoked by the make system, and that was mentioned in the original
> proposal http://mail-index.netbsd.org/tech-pkg/2005/03/20/0023.html
> about a year ago.
If you're talking only about the case when pkg_add is invoked by make
(and therefore you have the pkgsrc tree around anyway), then I withdraw
Gillette - the best a man can forget