Re: RFC: Alternate patch to have true new-style rc.d scripts in ports (without touching localpkg)

Mike Makonnen wrote:

> [...]
> Ok, can you do the following then:
>>>
>>> 1. When you (portmgr) are ready put back the rc.d/localpkg changes
>>> 2. Put the ordering of ports scripts with base system
>>>   scripts behind an rc.conf(5) knob, and modify your patch so both
>>>   /etc/rc and /etc/rc.d/localpkg do the right thing depending on
>>> whether
>>>   it's on or off.
>>
>> I suggest changing the extension for sourcing scripts to `.rc' and
>> ignore `.sh' scripts in rc/rc.shutdown.    The unmodified localpkg
>> should handle these.
>
> As I have already said, this is a gratuitous digression to support
> buggy rc.d ports script who's bugginess has only existed on
> 5-CURRENT up to now.

New features might break existing stuff, which doesn't imply that have 
been buggy. Usually you do an estimation of effort versus gain (at least 
that is what I usually do), and decide what is more promising to bring 
you forward.

I made two proposals:

Either

- allow new ports scripts to participate in rcorder(8), but disallow 
sourcing of scripts in /usr/local/etc/rc.d. Keep the historical 
difference that `.sh' scripts in /etc/rc.d are sourced, while `.sh' 
scripts in /usr/local/etc/rc.d are treated as old-style. No fixing will 
be necessary, but ports have to be changed to use the new functionality, 
supported by code in bsd.port.mk. Old or installed packages continue to 
work as usual.

or

- allow new ports scripts to participate in rcorder(8), and change the 
extension for sourcing of scripts to `.rc'. All `.sh' scripts will be 
treated as old-style, no matter whether they are in /etc/rc.d or 
/usr/local/etc/rc.d. Two scripts in /etc/rc.d (early.sh and rcconf.sh) 
have to be renamed, but no further fixing is necessary. Ports have to be 
changed to use the new functionality, supported by code in bsd.port.mk. 
Old or installed packages continue to work as usual.


As far as I understand your proposals are:

- put the localpkg commit back (which does rcorder(8) in ports, but not 
system-wide, so starting ports services early is still not possible) and 
let portmgr deal with the breakage

- source rc.d `.sh' conditionally on a configuration parameter in either 
/etc/rc or localpkg, and do rcorder(8) depending on that parameter. ~ 
100 ports have to be fixed and tested so that they will work correctly 
with the new semantics, and 5.0, 5.1, 5.2 and 5.2.1 are required to 
upgrade. New systems can not use old packages and have to upgrade 
installed ones.

> Since, you and I don't seem to agree on this issue and since we both
> don't seem to be restating the same things over and over againg.
> why don't we refrain for replying to this thread unless we have 
> something
> new to add?

Perhaps it would be beneficial if we summarize our proposals, so that we 
can have something like a ballot or ask re_at_ for their opinion. You might 
want to correct my summary of your proposals, since I'm sure I 
misunderstood some of your arguments.

-Oliver