ports/75018: [maintainer update]: net-mgmt/rotorouter was missing a file from its plist...
Paul Chvostek
paul+fbsd at it.ca
Tue Dec 14 21:20:34 UTC 2004
The following reply was made to PR ports/75018; it has been noted by GNATS.
From: Paul Chvostek <paul+fbsd at it.ca>
To: Sergey Matveychuk <sem at FreeBSD.org>
Cc: freebsd-gnats-submit at FreeBSD.org, ports at FreeBSD.org
Subject: Re: ports/75018: [maintainer update]: net-mgmt/rotorouter was missing a file from its plist...
Date: Tue, 14 Dec 2004 16:15:28 -0500
--FL5UXtIhxfXey3p5
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
On Tue, Dec 14, 2004 at 05:15:06AM +0300, Sergey Matveychuk wrote:
>
> I don't think it's a good idea to remove conf file this way.
>
> Use something like this instead:
> @unexec if cmp -s %D/etc/rotorouter.conf %D/etc/rotorouter.conf-example;
> then rm -f %D/etc/rotorouter.conf; fi
Agreed, that's more consistent with what other ports are doing. I've
attached a replacement diff against the original port.
But ...
While this is indeed consistent with how other ports work, it brings up
a philosophical question. Should port behaviour upon uninstall be
controlled by the port, or by the ports system itself?
By simply including the conf file in the +CONTENTS, without this
@unexec, a `pkg_delete` will remove the package but leave an altered
conf file, while `make deinstall` will remove the conf file as well.
Both will display a warning about the MD5 not matching, and neither will
return an error, whatever happens.
Could this be improved? I think so. (I'm CCing ports@ for discussion.)
So ... we have a "convention" amongst other ports of comparing config
files against their "-dist" or "-example" counterparts. Should this
behaviour be made standard? Perhaps a variable in the Makefile (i.e.
SAVEFILES or DO_NOT_DELETE) that contains a list of prefix-less files
that should kept rather than deleted if their MD5 is wrong?
For basic implementation of this, we wouldn't need to change the pkg_*
tools; the existence of the file in the variable could cause bsd.port.mk
to generate exactly the @unexec you've recommended above. Unless of
course we wanted some kind of override option that would let us say "ya,
delete the bugger, I don't care."
Is there any merit to any of this, or should folks installing ports just
be able to clean up after themselves?
p
--
Paul Chvostek <paul at it.ca>
Operations / Abuse / Whatever
it.canada, hosting and development http://www.it.ca/
--FL5UXtIhxfXey3p5
Content-Type: text/plain; charset=us-ascii
Content-Disposition: attachment; filename="rotorouter-1.0_1.diff-2"
diff -ruN /usr/ports/net-mgmt/rotorouter/pkg-plist rotorouter/pkg-plist
--- /usr/ports/net-mgmt/rotorouter/pkg-plist Mon Jun 14 13:56:41 2004
+++ rotorouter/pkg-plist Tue Dec 14 16:11:03 2004
@@ -1,3 +1,4 @@
sbin/rotorouter
etc/rc.d/rotorouter.sh
etc/rotorouter.conf-example
+ at unexec if cmp -s %D/etc/rotorouter.conf %D/etc/rotorouter.conf-example; then rm -f %D/etc/rotorouter.conf; fi
--FL5UXtIhxfXey3p5--
More information about the freebsd-ports-bugs
mailing list