Re: Config::General 2.31

On Sun, 2006-02-26 at 11:19 +0100, Terje Bless wrote:
> Ville Skyttä <ville.skytta@iki.fi> wrote:
> 
> >The main issue is just that as far as I can tell, Config::General 2.31
> >started behaving according to its own documentation regarding
> >-MergeDuplicateBlocks.  This is not something that should need changing
> >in the future nor needs to be documented IMO.
> 
> We have conditionals in our source for two versions of an external library;
> that's the test for a workaround that we drop and replace with “use Module
> <version>” once we find that appropriate (i.e. when Config::General >= 2.31
> seems a reasonable minimum prerequisite to set).

Given that the workaround is so trivial, I don't see a need for upping
the minimum required version at this point.  And especially so because
there's no version of C::G available that wouldn't need any workarounds.

The error message numbers in error_messages.cfg seem to be matched
against the numbers reported by onsgmls, so I gather working around this
issue by ensuring that the smallest message number in the config file
would be 1 instead of 0 would be possible but more obfuscated than the
current workaround.

> And as for documenting it, adding the archive URL for your message in the
> comment in “check” might be sufficient.

Ok, will add.

> And I still don't understand whether the workarounds apply to a bug in
> Config::General,

They do.

>  and if so, whether the bug was in earlier versions or was
> introduced in 2.31;

All versions, but it manifests itself a bit differently in < 2.31 and =
2.31.

>  or whether this is due a bug in the code in “check”, or
> something weird we're doing in the config files.

Nothing wrong with "check"; the problem is how Config::General handles
the "0" in <msg 0>, and I don't think that qualifies as weird.

Received on Sunday, 26 February 2006 11:31:02 UTC