[cvsnt] Re: info -b segfault regression between 2.5.03.2330 and 2344

[David Somers]

Bo Berglund wrote:
> I tested info -b on my Windows XP installation with build 2344 and
> there were no problems, except it found only 2 out of 3 CVSNT servers
> on my network. The one missing is a 2344 running on a VirtualPC W2003
> Enterprise server. Don't know why it is missing though....
> No segfaults or anything like that.

I've always found browsing and advertising has worked just fine for my
Windows-based servers and clients... but on the Linux side its been a tad
unstable.

TBH, I've never really managed to get cvsnt to advertise itself, except for
a few times back in the past when I used HOWL [which is now no longer
maintained]; issues seemed to arise when I migrated to Apple's
Bonjour/Rendezvous system. In the end I simply gave up and hacked xinetd so
that advertised cvsnt's presence instead (and as a side-effect, advertised
*all* services that were ran through it). Interested parties can see
http://www.omz13.com/tag/xinetd for more info on this hackery. Personally I
prefer xinetd to do the advertising rather than a service itself.

> cvs info itself also returns valid data with a cvsroot pointing to the
> missing server.

When doing cvs info -b, DNS Discovery is used to find the available hosts...
cvs info -r <host> talks directly to the host (using cvsnt's info
protocol)... which is why its possible to get the info from a host that
isn't advertising itself.

Greetings from (almost sunny) Luxembourg.

David Somers
programmer/typographer/whatever