Mailing-List: contact cygwin-developers-help AT sourceware DOT cygnus DOT com; run by ezmlm List-Subscribe: List-Archive: List-Post: List-Help: , Sender: cygwin-developers-owner AT sources DOT redhat DOT com Delivered-To: mailing list cygwin-developers AT sources DOT redhat DOT com Date: Wed, 20 Jun 2001 11:42:01 EDT To: cygwin-developers AT cygwin DOT com Subject: Re: Comments on Robert's category feature X-Mailer: Virtual Access by Atlantic Coast PLC, http://www.atlantic-coast.com/va Message-Id: From: Brian Keener Reply-To: bkeener AT thesoftwaresource DOT com In-Reply-To: References: Robert Collins wrote: > Sure. I agree with you. The point about "installation methods" is that > they _require_ dependencies to operate. They don't _require_ package > categorisation. The reason they require dependencies is that you are > specifying a list of packages that need to be installed. And I agree with you - all true. I was just voicing what my original thought/interpretation of where we were headed with setup. The only exception I have to what you said is that dependencies are not a *prerequisite* to installation methods. Each package could be coded with an installation method(s) in setup.ini just as they could be coded with category or categories and these could be parsed just as things are now. The true need for dependencies comes into play when the user begins altering the initial default selections that were made by the setup program as it then becomes necessary to check they have not selected a conflicting package or deselected a package needed by some other package. I am by no means advocating that the current development process is wrong or should be changed, I am only clarifying what I perceive the process and different options to be. > > On a side note as well I tried to try the dependecies logic > > using an update > > from cvs and Roberts sample setup.ini - I don't see anything > > different - where > > is the categories. > > You will need chris's updated copy of my patch before anything happens > to the setup.exe interface. The code in CVS is just the parser, not the > operational logic. > I think I downloaded that patch - was that the one Chris posted with the subject 'Updated Robert Collins dependency/category patch against current CVS' and the comment that he felt bad about changing code around your patch code and had rerun the diffs against the current CVS. I'm currently trying to get it to patch - my email program mucked with the CR/LF and the tabs. bk