On Feb 20, 2014, at 4:20 PM, Gerald Combs <gerald@xxxxxxxxxxxxx> wrote:
> I'd prefer to have as few repositories and repository URLs as possible.
> This would hopefully make things less confusing for developers and is
> one less thing to monitor and maintain on the server side. (On the other
> hand I think there's a compelling argument for mirroring to GitHub,
As long as all mirroring either
1) treats the mirror as read-only
or
2) something automatically pulls changes from the mirror to the Official Repository
and
3) github doesn't become a replacement for Bugzilla.
I speak from experience with libpcap and tcpdump, which:
have both repositories on bpf.tcpdump.org and GitHub; changes checked into bpf get pushed to GitHub automatically (I don't know how often - it might be nightly) but changes on GitHub (such as merged pull requests) currently require that they be *manually* pulled (I have a "pull-from-github" script that lets me do that to my libpcap and tcpdump trees, and then I push them to bpf);
use GitHub as the bug database - meaning it doesn't support attachments.
> (It looks like people are still checking out from
> anonsvn.wireshark.org/wireshark. I'm not sure what to do about that.)
Make one last checkin to it that changes autogen.sh to just print
THESE AREN'T THE WIRESHARK SOURCES YOU'RE LOOKING FOR. MOVE ALONG TO {url}
and exit with an exit status of 1, and perhaps tweak the configure script to do the same?