PATCH: hgweb_dir.py - Work in proxied environments

[Matt Mackall]

On Mon, 2010-08-16 at 15:19 -0400, Michael Tharp wrote:
> On 08/16/2010 10:14 AM, Dirkjan Ochtman wrote:
> > On Mon, Aug 16, 2010 at 16:21, Steve Kemp<steve at steve.org.uk>  wrote:
> >>   I see no downside to using HTTP_HOST rather than SERVER_NAME, so the patch
> >>   below makes that change.
> >
> > Me neither, so I'm inclined to take this. Anyone know any disadvantage?
> 
> The Host header includes any port supplied by the client, so adding one 
> again will cause problems if something other than 80/443 is used.
> 
> e.g. http://foobar/ -> foobar, http://foobar:8080/ -> foobar:8080
> 
> Corrected patch below. Sorry in advance if it gets mangled.
> 
> -- m. tharp
> 
> 
> diff --git a/mercurial/hgweb/hgweb_mod.py b/mercurial/hgweb/hgweb_mod.py
> --- a/mercurial/hgweb/hgweb_mod.py
> +++ b/mercurial/hgweb/hgweb_mod.py
> @@ -218,7 +218,7 @@
> 
>           port = req.env["SERVER_PORT"]
>           port = port != default_port and (":" + port) or ""
> -        urlbase = '%s://%s%s' % (proto, req.env['SERVER_NAME'], port)
> +        urlbase = '%s://%s' % (proto, req.env['HTTP_HOST'])
>           staticurl = self.config("web", "staticurl") or req.url + 'static/'
>           if not staticurl.endswith('/'):
>               staticurl += '/'

My understanding is that HTTP_HOST isn't always available. In
particular, with HTTP 1.0 clients. Not sure if that matters any more.

-- 
Mathematics is the supreme nostalgia of our time.