Glad to be a guinea pig, @bradrydzewski! I fired up this new image in my clone phase, and got this error right out-of-the gate on the /pr job (/push job ran normally):
chmod: 600: No such file or directory
Initialized empty Git repository in /drone/src/github.com/myorg/dummy-repo/.git/
+ git fetch origin +refs/heads/master:
From https://github.com/myorg/dummy-repo
* branch master -> FETCH_HEAD
* [new branch] master -> origin/master
+ git checkout master
Branch master set up to track remote branch master from origin.
Already on 'master'
+ git fetch origin pull/28/head:
From https://github.com/myorg/dummy-repo
* branch refs/pull/28/head -> FETCH_HEAD
+ git rebase 38482f303c72888a96d50338a5d8b466db9c097d
First, rewinding head to replay your work on top of it...
*** Please tell me who you are.
Run
git config --global user.email "[email protected]"
git config --global user.name "Your Name"
to set your account's default identity.
Omit --global to set the identity only in this repository.
fatal: unable to auto-detect email address (got '[email protected](none)')
Hereās my drone.yml:
clone:
git:
image: plugins/git:next
pipeline:
canary:
image: alpine
commands:
- echo "This is a test to make sure Drone is running jobs."
I added a default email and name (via environment variable) which should satisfy that github prompt if it appears again. A new image is now available. Note that I didnāt have a test PR to reproduce the prompt, so this is untested, but according to git docs it should work.
Thanks - Iāll try this. Since I already have the :next version installed on several agent machines, is there an easy way to force the upgrade without manually removing that image from each machine?
First of all, big thanks for this. Outdated merge heads from Github started to pop up more and more in our deployment.
We now started using this in a few repo. First executions look good, and we are happy that this is hash based now.
There was one case though that may be of interest: in one of our long running branches the rebase failed with a conflict, while a manual merge of master to this branch succeeded without conflict. Retrying the build after this merge, the rebase succeeded as nothing left to rebase really.
Iām thinking that git rebase <hash> is perhaps more fragile than git reset --hard <hash> followed by a git merge master. But Iām sure you have reasons why itās a rebase and not a merge.
not really, it was just the first thing that passed unit tests
The primary goal was was to preserve the hash and not create a new merge commit or create a detached head state. If the unit tests pass without modification with reset --hard I am open to adjusting. Thanks for testing, and for the feedback!
Just wanted to verify that weāll be able to continue to pull the specific branch/commit without doing any of the auto-rebasing stuff. In our case, the extra magic hurts us more than it helps.
@gtaylor I recommend either testing the shell script locally, or maybe even testing with one of your repositories if possible. If we identify issues over the next few weeks, it will give us time to fix before this becomes the default in 0.9
With those criteria I couldnāt come up with an alternative. My original proposal creates both a detached head (since we donāt know the source branch) and a merge commit. So I stick with whatās there already. Thanks.
Itās very minimal feedback Brad, but ānextā is not picking up the fact weāre pulling from GitLab and changing the ref path. (Drone : 0.8.2). I assume thatās because the $DRONE_COMMIT_REF isnāt being evaluated correctly - but Iām not sure how to make that obvious and get that information for you. Let me know if I can be of any help.
Update: Hereās what I get from my build (and I donāt know why it doesnāt pick it up in ānextā)
echo $DRONE_COMMIT_REF
refs/merge-requests/1/head
Update: Seems itās down to the version of /bin/sh. Iāve submitted a PR with a workaround.
I tried using git:next and it seems that cd to $DRONE_WORKSPACE fails since the directory does not exist.
On the other hand, it seems that code in master branch creates the workspace directory if it does not exist.
It would be helpful if you could add the same procedure (create the workspace directory) to next branch.
Here is a data point - the GitHub refspecs have been particularly flaky today, meaning that Drone was giving very confusing errors for hours. I ended up switching that repo to plugins/git:next, and the problems disappeared immediately. Weāre going to merge it into master, as I donāt see any regression or bug in that version
Weāve been using it and it has been great. I would love to see it merged because it is a pain to have to add it. When I forget it we have issues on new projects.
@bradrydzewski Thanks for working on improving the cloning process! We are running into issues with drone building non-current versions of PRs on a weekly basis.
Having said that, I would like to suggest a) using git merge 26923... instead of git rebase 26923... since this more closely resembles what happens on Github when you merge a PR. Also, there can be merge conflicts for a rebase that do not exist for a merge.
And b) the whole idea of āwe merge ourselves because Github is too slowā is not really implemented currently, since with git fetch origin pull/14596/head: we still depend on the Github pull ref to be up-to-date. And because I see now way of avoiding to use the Github pull ref for inter-repository PRs, I would suggest to simply check that the fetched ref matches the DRONE_COMMIT_SHA and retry if not (yet).
If these ideas are welcome, I would volunteer to work on a PR for drone-git:next for this!?
Or should it be a PR for drone-git:master for only my idea b)?
Were there other reasons for next except outdated Github pull refs? (Yes, I have seen the listed benefits at the top post, butā¦ 1) is true but maybe not so relevant since github itself tells you something about merge failures already, 2) is also in my idea b), 3) is not possible for inter-repo PRs and 4) I do not really understand - since the head sha will be changed either way (true for rebase, reset and merge)
Thanks for the feedback. The window for making changes has expired. The current implementation is being shipped in 0.9 and we have a number of teams using the new version successfully.
I suggest creating a custom git plugin and sharing with the community. This gives the community a chance to test your plugin (without the risk of breaking changes) and then suggest upstream improvements after successful testing.
I also think there might be some flaws in your design / assumption. I started writing a critique but it was taking a while, and I do not have the free time to allocate to this right now ā but I do think that creating your own custom plugin and using it for a few weeks, and getting others to use it, would be a helpful exercise and would likely surface some issues and allow you to tweak the design.
The current implementation is being shipped in 0.9
That means we can think about changes for 0.10 or 1.0
I started writing a critique but it was taking a while, and I do not have the free time to allocate to this right now
I totally get that! It is a complicated topic and I am sure I am also missing some corner cases! If you want, have a quick look at the following argument to support my suggestion a):
This shows a branch fourth conflicting with a commit (āthird branch conflicting with secondā) inside the master branch but not with the HEAD of master:
Initialized empty Git repository in /drone/src/github.com/ploh/dronetest/.git/
+ git fetch origin +refs/heads/master:
From https://github.com/ploh/dronetest
* branch master -> FETCH_HEAD
* [new branch] master -> origin/master
+ git checkout master
Branch master set up to track remote branch master from origin.
Already on 'master'
+ git fetch origin pull/3/head:
From https://github.com/ploh/dronetest
* branch refs/pull/3/head -> FETCH_HEAD
+ git rebase be6ba555aabaaa422cdbe2a62a083b36ae28419e
First, rewinding head to replay your work on top of it...
Applying: third branch conflicting with second
error: Failed to merge in the changes.
Using index info to reconstruct a base tree...
M .drone.yml
Falling back to patching base and 3-way merge...
Auto-merging .drone.yml
CONFLICT (content): Merge conflict in .drone.yml
Patch failed at 0001 third branch conflicting with second
The copy of the patch that failed is found in: .git/rebase-apply/patch
On the other hand, I have not heard an argument for using rebase (and even in the āwrongā direction, since we are rebasing master onto PR branch with git:next) instead of merge. I will create a custom plugin with this but in the meantime I just wanted to make clear why this matters
On the other hand, I have not heard an argument for using rebase
It is mentioned in my initial post, but rebase preserves the head sha, which is important for coveralls, codecov, etc. Without a rebase it creates a new merge commit, which breaks these tools, and could also cause problems when we eventually try to support the github checks api.
But rebase will also create new commits. It will rebase master on top of the PR branchās HEAD. Whereas merge will create a merge commit on top of the PR branchās (as well as masterās) HEADs. Either way we change the commit sha of the pushed PR branch. That is unavoidable imo.
btw: Thanks for taking some time to try to enlighten me - even if unssuccesful until now
