Call for testers -- CONF_FILES variable

[Chris Rees]

On 14 July 2011 20:02, Eitan Adler <lists at eitanadler.com> wrote:
>> The reason I choose pkgconf (we can change that name) is that it concerns
>> only configuration files that the maintainers DO want.
>>
>> I want to make sure that maintainers are looking at the samples the proprose
>> to provide a usable sample, not the default one from the distfile (the
>> default one can still be provided as an example.)
>
> So, this suffix is only for configuration files that port maintainers
> write and included sample files from upstream
> would not have this suffix?
>
> Why would the maintainer be writing sample conf files? It is not the
> maintainer's job to write documentation for the upstream project. The
> only case I could see this becoming an issue is if the default
> configuration file ignores hier(1) and a REINPLACE is needed. I do
> _not_ want to see sample configuration files being written for ports
> unless a considerable amount of rework is needed to make the
> application run on FreeBSD.

What bapt is talking about is that he doesn't want people to blindly
install the .sample files from the distfile, and actually _look_
through them.

Of course, if you're changing the files at all you really shouldn't
use the .sample format, because the .sample format comes from the
distfile, not necessarily the port.

Being explicit about .pkgconf (or whatever colour it is) shows that
the maintainer is responsible for the sample config file rather than
the upstream.

I think it's much politer for the users to receive a config file
that's almost usable.

>> I wanted that pkgng and the ports in general can manage default usable
>> configuration files, and to distinguish them from the samples. Thanks crees@
>> has done the job I wanted to do myself so that and he has done it right.
>
> I am confused. I thought Chris's option was for the upstream sample
> configuration files.My understanding is that it replaces the logic of
> "only delete the real config file if it does not differ from the
> sample file".  Why then does it matter who wrote the sample since the
> logic works the same way? Either we will need multiple copies of this
> macro, one for "official" files and the other for "package" files or
> the logic will still have to be replicated per port for non-included
> samples. IMHO the suffix (and type of sample file) should not be
> touched by the macro.
>
> Perhaps I misunderstand what will be new in pkgng or what this patch provides?
>


Chris