Submit repo status (original) (raw)

Aleksey Shipilev shade at redhat.com
Tue Dec 11 13:05:34 UTC 2018

On 12/10/18 9:49 AM, Volker Simonis wrote:

The problem you'll run into when you create new clones from time to time is that a single push from somebody who didn't realize that you have created a new clone (and didn't recreate its local version of submit) can unintentionally push ALL the old branches from his local clone when he only wants to push his single, new branch. I actually wonder that this hasn't happened already ...

+1.

I was staring blankly at my current "hg out" for jdk-submit trying to understand why does it want to push thousands of changesets out. Only to realize something is "wrong" with the remote repository. Then I discovered this thread. I could have easily not check it and push.

I have three suggestions for the future: a) Please say "jdk-submit is reset, please re-clone" as the message synopsis, not the innocuous "Submit repo status" -- this important message was easy to miss; b) Test if pushing old changesets accidentally is caught by some checker. I wonder if current single-branch checkers would still allow to push all old changesets; c) Support clonebundles [1]! It took me five (!) re-tries to clone new jdk-submit for my workspace cache [2]. It would be very unfortunate to go through this exercise regularly.

Thanks, -Aleksey

[1] https://bugs.openjdk.java.net/browse/JDK-8211383 [2] https://builds.shipilev.net/workspaces/

More information about the jdk-dev mailing list