There are certain cases in chrome-os where we bypass review, and do direct
uploads. Things like builders signing off on a change, thus us up-rev'ing a
manifest/ebuild to expose the actual change to our stable users.
For this, we've basically hardcoded the server to push to; this however sucks
since we have the necessary info in the manifest already (and hardcoding breaks
down if two review servers are in use).
As such, attached is a patch injecting a pushurl if a review url is specified
for the remote.
The flaw with this is that if the gerrit server's ssh_info has changed since
the user sync'd, the pushurl will still point to the old location; there really
isn't any way around that if we're exposing `git push`.
If that's acceptable, ok, patch is attached. If not, I can hack up a repo
push; I'm wary to do this however since repo push isn't exactly equivalent to
`repo upload`; uploading several feature branches in parallel is sensical.
Trying to push multiple feature branches at the same time however is a bit
dodgy- the user has to ensure they will merge nicely (which if not, means one
is already live).
Frankly I don't like that; could do repo push branch [project]+ instead (thus
preclude the issue via limiting the UI to single branch pushing).
Since most folk are using uploading, I went the git config route instead. If
that doesn't suffice let me know and I'll hack out the subcommand approach.
Original issue reported on code.google.com by ferri...@chromium.org on 22 Dec 2011 at 7:39
Original issue reported on code.google.com by
ferri...@chromium.org
on 22 Dec 2011 at 7:39