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

[Oliver Eikemeier]

Alex Kapranoff wrote:

> * Oliver Eikemeier <eikemeier at fillmore-labs.com> [July 29 2004, 12:39]:
>> I'm not sure if we should filter out *all* scripts with extensions.
>> There might be a startup.app port we like to add, perhaps just a list 
>> of
>> values like `.old', `.sample' e.t.c. should be filtered out. OTOH this
>> is easily changeable, should the need arise.
>
>   And startup.app port can easily be started with startup_app script,
> This is a non-issue, I think.

Jup, I'm thinking in terms of complexity and POLA here, not 
functionality.

>> which will install `apache' or `apache.sh' depending on OSVERSION. A
>> variable RC_SUFX is set that could be used in pkg-message or other
>> places when necessary.
>
>   The need for two versions of startup scripts for each port is highly
> undesirable. There should be some shims for 4.x systems, I suppose,
> for them to be able to execute extensionless rc.d scripts in simple
> lexicographic order (and missing all the rcorder benefits).

No, every port just has one version of a startup script, it is just 
*installed* under a different name, depending on OSVERSION. The 
borderline is not 4.x/5.x, it is before/after the patch (which enables 
execution of extensionless rc.d scripts). The supporting code in 
bsd.port.mk could install shims that enables newer 5.x packages to be 
used on older systems, but I would prefer not to have them.

>> Sourcing port scripts is not possible with this patch, which is a good
>> thing IMHO. Also some documentation needs to added to rc(8) before this
>> patch can go in.
>
>   I like the patch and would like it to go into the tree instead of
> Mike Makonnen's solution. Sourcing 3rd party (ports) scripts into the 
> shell
> which performs global system startup procedures worries me.

Thanks. Mike's patch sources scripts only into localpkg, so there is an 
additional layer, yet they can influence each other (by erroneously 
setting common variables). I can not see what sourcing ports scripts 
should be good for, though.

-Oliver