[Savannah-hackers] Re: commit_notify

[Jaime E . Villate]

On Tue, Apr 17, 2001 at 10:01:01AM +0200, address@hidden wrote:
> I feel bad, that's obviously the most complex part of it and
> I think I did not specify this to you. Very sorry. Each project with
> a source repository has its own CVSROOT in /cvsroot/project. The 
> /cvsroot is also a CVSROOT but is only used for anonymous access. 
> If you look in /etc/xinetd.conf you'll see that the cvs pserver is called
> with a huge amount of cvsroot to solve this problem. 
>       I guess viewcvs must be patched to allow that kind of setup in
> a way that is not as cumbersome. 

I had read that already in "savannah.html" before I configured viewcvs, and
there is no problem from the point of view of viewcvs; since each project's
repository is a subdirectory of /cvsroot/, viewcvs has been configured to
see only the main repository in /cvsroot and it accesses the project's
repositories as if they were subdirectories of a single project.

I believe I have fixed the problem with the URL's in the e-mail
notifications. Since "log_accum" is given the project's name with the -T
option, I use that name (when necessary) to construct the right path to the
project's repository. The -T option of log_accum formerly meant "temporary
file name", but from now on it means "project's name" (which happens to be
used as temporary file name too).

Now I have a more serious problem for which I need some help from those of you
more familiar with subversions before Savannah started. cvsweb was installed
differently in subversions and savannah. For example, if the same repository
"gnudocs" is accessed in the following two ways:
  http://subversions.gnu.org/cgi-bin/cvsweb/gnudocs/
  http://savannah.gnu.org/cgi-bin/cvsweb/gnudocs/

The first one shows the content of the directory, but the second one fails
because /cvs/gnudocs is not publicly available. In the first case cvsweb
points to a wrapper that will give it some root privileges; in the second case
cvsweb is run directly off /usr/lib/cgi-bin/ and the userid of the httpd
server is not allowed to access /cvs/gnudocs

What should I do with cvsview? do we still want a wrapper as the one cvsweb
had? (I guess with ssh cvs access there is no need for it) and should
 http://subversions.gnu.org/cgi-bin/viewcvs/ exist or is it enough with
http://savannah.gnu.org/cgi-bin/viewcvs/?

cheers,
Jaime