Re: [PATCH] Remove dependency on IO::String from Git.pm test

From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Jun 18, 2008, 19:35 UTC
Message-ID: <alpine.DEB.1.00.0806182030200.6439@racer>
In-Reply-To: <200806181952.02048.jnareb@gmail.com>

Hi,

On Wed, 18 Jun 2008, Jakub Narebski wrote:

Show 6 quoted lines

> On Wed, 18 Jun 2008, Johannes Schindelin wrote:
>
> > BTW I think it is not nice at all how the dependency hell with Git.pm 
> > is made worse recently.
> 
> It is not dependency for Git.pm, but for Git.pm TEST.

So? Why do you want to break the _test_ on those machines where you need them most? We _know_ that a release from Junio works fine on Linux.

Show 7 quoted lines

> > It is fascinating through how much _pain_ we go with the shell scripts 
> > to maintain portability, even with _very_ old or obscure systems (see 
> > the SCO server patches that came in not long ago!), and just walk over 
> > that portability when it comes to Perl...
> 
> And I pointed out how it could be resolved (use 5.8 specific feature, or 
> IO::String, or skip tests).

I have to point out that the platforms I was speaking of are not know to make upgrading as easy as Linux. And some of them _do_ come with pretty old perl. And yes, I had this exact issue (remember when I worked on removing Git's dependency on Python? That was it. Not enough quota. Uncooperative admin. Desperate need for a sensible SCM).

In any case, I have to reiterate my point: breaking backwards-compatibility for _no_ good reason is wrong. I have not looked at the patch in question, but I seriously doubt that this is the easiest, most elegant (and yes, backwards-compatible) solution.

Typically, it is not a good sign when you _require_ newer and newer features and versions of components you use.

Whatever, Dscho