Re: Test 17 of `t9119-git-svn-info.sh' fails with svn 1.6.9

From: Frank Terbeck <ft@bewatermyfriend.org>
Date: Jul 1, 2010, 10:36 UTC
Message-ID: <87lj9vebfx.fsf@ft.bewatermyfriend.org>
In-Reply-To: <4C2C6D64.4060802@drmicha.warpmail.net>

Michael J Gruber wrote:

> Frank Terbeck venit, vidit, dixit 01.07.2010 12:19:
>> Michael J Gruber wrote:

[...]

Show 12 quoted lines

>>> Do the following agree for you:
>>>
>>> git svn --version
>>> svn --version
>>>
>>> Do you have any private copies of svn/svn bindings in your path or perl
>>> path?
>> 
>> Okay, it wasn't really a private copy, but you're otherwise
>> spot-on. Here's what was wrong:
>
> Turns out you cc'ed the right persons :)

Heh, seems so. :)

Show 15 quoted lines

>> I didn't realise, `git-svn' was using Perl's subversion bindings. This
>> happens on my laptop which runs debian stable (lenny). I was upgrading
>> my subversion package to the one provided by backports.org (a service
>> that provides backports of newer versions of certain software packages
>> to debian stable). While that worked nicely, the subversion bindings for
>> Perl are provided by a package called `libsvn-perl' which was kept at
>> the version in stable, which is 1.5.1. Backports.org also provides an
>> upgrade for that package. Now both "svn --version" and "git svn
>> --version" report the same subversion version and the test passes again.
>> 
>> 
>> Maybe it would be good if git-svn or maybe just the test suite checked
>> whether subversion's and the Perl binding's version matched?
>
> Then you would not be able to use git-svn with your setup!

Well, now it would. :) But I get your point.

> Having svn and its bindings at different versions is perfectly fine for
> git-svn: it uses and cares about the bindings only (and adjusts
> according to the binding's version).

Okay.

Show 5 quoted lines

> It's only the test suite which cares, because it tries to make sure that
> "svn" and "git-svn" behave as similar as possible.
>
> We should probably check in the test suite before triggering false
> alarms, I'll look into that.

Cool thanks.

And thank you for the quick response and solution.

Regards, Frank