Re: [egit-dev] egit.eclipse.org now running smart HTTP

["Shawn O. Pearce" <spearce@xxxxxxxxxxx>]

Matthias Sohn <matthias.sohn@xxxxxxxxxxxxxx> wrote:
> 
> I tried to clone using the HTTP URL given by gitweb
> http://egit.eclipse.org/r/p/jgit.git
>        $ git clone http://egit.eclipse.org/r/p/jgit.git
>        Initialized empty git repository in ...
>        warning: You appear to have cloned an empty repository.

Something is wrong with the Apache reverse proxy in front of Gerrit.
If I pull back the reference list for the repository its wrong:

  $ curl -I http://egit.eclipse.org/r/p/jgit.git/info/refs?service=git-upload-pack
  HTTP/1.1 200 OK
  Date: Tue, 19 Jan 2010 07:21:46 GMT
  Server: Apache/2.2.0 (Linux/SUSE)
  Content-Type: text/plain; charset=ISO-8859-1


But when I login to egit.eclipse.org and hit the Gerrit process
directly its correct:

  $ curl -I http://localhost:8081/r/p/jgit.git/info/refs?service=git-upload-pack
  HTTP/1.1 200 OK
  Date: Tue, 19 Jan 2010 07:21:27 GMT
  Expires: Fri, 01 Jan 1980 00:00:00 GMT
  Pragma: no-cache
  Cache-Control: no-cache, max-age=0, must-revalidate
  Content-Type: application/x-git-upload-pack-advertisement
  ETag: b43b199cb1dbb8c1c6b7803ddc7688f5527570c7
  Content-Length: 25292

So Apache is doing something odd here and just destroying the
response.  No wonder its not working.

Oh, I know what it is.  Its a pattern match on /.*\.git/.* that is
bypassing the proxy into Gerrit.  I'll fix the configuration file
and restart the server in a minute.
 
>        $ git clone http://egit.eclipse.org/r/p/jgit
>        error: the requested URL returned error: 403 while accessing
> http://egit.eclipse.org/r/p/jgit/info/refs

Odd, this just worked for me:

  $ git clone http://egit.eclipse.org/r/p/jgit
  Initialized empty Git repository in /Users/sop/tmp/jgit/.git/
  remote: Counting objects: 2175, done
  remote: Compressing objects: 100% (2175/2175)
  Receiving objects:  52% (1131/2175), 131.41 KiB | 19 KiB/s   

Can you retry with the enviroment variable GIT_CURL_VERBOSE=1 ?
 
> but the following call succeeds :
>        $ git clone http://egit.eclipse.org/jgit

This is going through the dumb file server in Apache, and is *not*
using the smart HTTP interface.
 
-- 
Shawn.