Re: [PATCH v2] git-cvsimport: add support for cvs pserver password scrambling.

From: Dirk Hörner <dirker@gmail.com>
Date: Aug 14, 2009, 09:46 UTC
Message-ID: <4da546dc0908140246i35b28052wd7b7790d191a9984@mail.gmail.com>
In-Reply-To: <7vskfusvpq.fsf@alter.siamese.dyndns.org>

Hi Junio,

On Fri, Aug 14, 2009 at 9:25 AM, Junio C Hamano<gitster@pobox.com> wrote:

Show 17 quoted lines

> Thanks.
>
> While I appreciate your effort to add a test, I'd rather not apply the
> test part of your patch for two reasons:
>
>  - It is not a test against a real cvs pserver but is a whitebox test to
>   verify that the program says what the program is supposed to spit out
>   to the network; and
>
>  - It still is a network test that will fail if the TCP port is occupied
>   for whatever reason when the test is run, which will make automated
>   build and test cycle unreliable.
>
> Unfortunately, I do not see an easy way to run a real cvs pserver
> listening to a local unix domain socket under $TRASH_DIRECTORY, which
> would solve both of the above issues.
>

I agree with you, the tests are not the best. As Dscho stated in one of the replies to the last version of this patch, the cvs pserver is quite good to test with because it uses stdin/stdout for communication. The main problem is git-cvsimport, which right now only supports tcp sockets.

Ciao, Dirk