Re: [PATCH] Use new .git/config for storing "origin" shortcut repository

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Nov 26, 2006, 13:45 UTC
Message-ID: <Pine.LNX.4.63.0611261437270.30004@wbgn013.biozentrum.uni-wuerzburg.de>
In-Reply-To: <7vzmaeyie6.fsf@assigned-by-dhcp.cox.net>

Hi,

On Sat, 25 Nov 2006, Junio C Hamano wrote:

Show 9 quoted lines

> Johannes Schindelin <Johannes.Schindelin@gmx.de> writes:
> 
> > This hunk is wrong: "git-repo-config remote.bla.fetch b" will _overwrite_ 
> > remoter.bla.fetch. To avoid that, you have to use "git repo-config 
> > remote.bla.fetch b ^$". (The last argument is a regular expression which 
> > has to be matched by the value-to-be-oreplaced.)
> 
> Yup, one of the joys of working the config file everybody seems to like 
> ;-).

The problem is: config files are ubiquitous, so you need not teach users about it. On the other hand, they are just key / value stores, i.e. reflecting a mapping. What we want here, is a multimapping, so we use the wrong tool.

But sometimes it is so much more pragmatic to just take off one shoe and put the darned nail back into the wall than to go to the shop, buy the hammer, go back, put the nail in, and try to sell the hammer via eBay.

BTW regarding your criticism of the config file: I agree that the write support of git-repo-config was quite brittle at the start. Which is my fault.

However, we had quite some flashing out bugs in the mean time, so I am quite confident in the tool. Of course, what with the recent addition of a user specific config file (which makes the name "repo-config" seem utterly wrong), there might be some dragons in the code.

So, it seems that the whole config writing code is a perfect opportunity for people wanting to audit source code!

Ciao, Dscho