Re: Bug with URL-decoding

From: Jeff King <peff@peff.net>
Date: May 23, 2010, 09:16 UTC
Message-ID: <20100523091612.GB26123@coredump.intra.peff.net>
In-Reply-To: <loom.20100507T000406-241@post.gmane.org>

On Thu, May 06, 2010 at 10:06:50PM +0000, Dave Abrahams wrote:

Show 9 quoted lines

> $ cd /tmp
> $ mkdir x+y
> $ cd x+y
> $ git init
> $ cd ..
> $ git clone file:///tmp/x%2By zz  # <==== ERROR
> $ git clone file:///tmp/x+y zz # <====== OK
> 
> Funny characters do come up in names, especially in test suites

Yeah, at this point file:// is more or less just an alias for a straight path (and ssh:// an alias for "host:path" syntax). We correctly percent-decode http:// URLs because we just hand them to curl, which does the right thing. We should be decoding these ones when we pick apart the URL components.

Patch series to follow:

  [1/2]: make url-related functions reusable
  [2/2]: decode file:// and ssh:// URLs

With these patches, you can do:

  git clone file:///tmp/x%2By

or

  git clone /tmp/x+y

The former will clone into "x%2By". Which I'm not even sure is wrong, so I left it as-is.

-Peff