"already installed" and other problems

[John Klos <john%ziaspace.com@localhost>]

A number of observed pkgsrc problems, in summary:

(1) when doing a "make update", the package being updated should NOT be 
deleted before the compile stage. Please explain to me why you disagree, 
if you do.

This would have jus incapacitated my mailman server if I had done a "make 
update" on python25 on my PowerPC server because libffi doesn't compile 
properly on PowerPC, so python would've been deleted, but then wouldn't be 
buildable. On a production server, I don't want to have to do a "make", 
then a "make update", then hunt down any new dependencies so that I can do 
a "make clean" in them, too.

(2) when doing a "make update" on a package which might be the 
prerequisite for tons of other packages (for example, perl), there are 
consistently problems with the process stopping with "already installed" 
errors even though there are definitely no stale work directories.

This problem is tremendously burdensome on busy systems which you can't 
exactly stop and which are heavily used. I'd like to be able to do a "make 
update" and NOT have to babysit the entire process. Even fast systems 
which have heavy loads can take a long time to do a full "make update" on 
perl.

If anyone thinks this is an isolated problem or a problem which only 
occurs on certain systems, I'd be happy to set up a NetBSD-5 machine from 
scratch with the latest pkgsrc, then install the common packages I use 
(Apache, BIND, php, perl, MySQL, et cetera), then do a "make update" in 
perl. It happens on many different kinds of systems.

(3) (small problem) automatically becoming root is inconsistent. It'd be 
nice just for consistency's sake. "make update" and friends are automatic, 
but pkg_delete and friends are not.

Just some observations...