From: Junio C Hamano <gitster@pobox.com>
Date: Wed, 08 Apr 2009 06:29:59 GMT
Subject: Re: [PATCH] perl: add new module Git::Config for cached 'git config' access
Message-ID: <7vbpr7r72w.fsf@gitster.siamese.dyndns.org>
In-Reply-To: <m3prfo1xh6.fsf@localhost.localdomain>

Jakub Narebski <jnareb@gmail.com> writes:

> Errr... wouldn't it be better to simply use 
>
> +		my ($item, $value) = split("\n", $_, 2)
>
> here? Have you tested Git::Config with a "null" value, i.e. something
> like
>
>     [section]
>         noval
>
> in the config file (which evaluates to 'true' with '--bool' option)?
> Because from what I remember from the discussion on the 
> "git config --null --list" format the lack of "\n" is used to
> distinguish between noval (which is equivalent to 'true'), and empty
> value (which is equivalent to 'false')
>
>     [boolean
>         noval        # equivalent to 'true'
>         empty1 =     # equivalent to 'false'
>         empty2 = ""  # equivalent to 'false'

I do not mind if the _write method always wrote out

	[core]
        	autocrlf = true

for a variable that is true, but it should be able to read existing

	[core]
        	autocrlf

correctly.

Sam, I think you meant to make me squash the "Oops, for no good reason,
here is a fix-up" into the previous one, but for this case, I'd appreciate
a re-roll of the series, that includes a test to read from an existing
configuration file that contains such "presense of the name alone means
boolean true" variables.