Hubert Feyrer wrote:
> On Tue, 6 Sep 2005, Matthias Drochner wrote:
> 
>> I'm always pissed off when someone renames package directories
>> forth and back (this breaks pkg_chk and automatic updates, let alone
>> local and uncommitted packages), so I'd prefer to turn math/gnumeric
>> into a meta package here.
> 
> 
> So fix pkg_chk.
> Esp. having one pkg pull in another one because we can't easily rename 
> pkg dirs seems wrong to me.

How do you feel about leaving a record of deprecated packages lying 
around, in order to a) make tools like pkg_chk work, and b) answer the 
inevitable "wtf?" question from users when their favourite package no 
longer seems to exist?

One way to do this would be to replace the old package Makefile with a 
Makefile containing something like the following:

DEPRECATED=renamed
DEPRECATED_NEWNAME=foo/bar
DEPRECATED_REASON=Keeping users on their toes

Possible values of DEPRECATED could include, for example: renamed, 
superseded, deleted

As a matter of policy, deprecated packages should be kept around for a 
while (2 branches/6 months?) and then simply deleted.

Attempting to build such a package should fail with a textual message 
incorporating the values of the DEPRECATED* variables.

pkg_chk should optionally either break with the above error message 
(current behaviour), or be smart enough to follow the link to the new 
package and build that instead. If following links, pkg_chk absolutely 
must list deprecated packages at the end of the run so that users know 
they must update their pkgchk.conf files.


This could also be done in a separate file, but something tells me that 
this way is probably easier to implement, and is more likely to be 
noticed and kept up to date by developers.

-- 
Gillette - the best a man can forget