[v4] Commit creating failed: ["Service not found: none"]

[dfed]

I'm trying to migrate to v4 and I've encountered two errors on an ubuntu-latest runner.

First: same error message as codecov/codecov-action#1252, which I fixed by adding a Setup Python step to ensure that 3.10 was installed.

Second: Commit creating failed: ["Service not found: none"]

I haven't yet found a workaround for this second issue.

For what it's worth, my macOS runner is succeeding with v4.

---

[@thomasrockhu-codecov edit]
The root cause of this error is that all git calls made by the CLI here are failing. The error message is

fatal: detected dubious ownership in repository at '/__w/example-python/example-python'\nTo add an exception for this directory, call:

    git config --global --add safe.directory /__w/example-python/example-python

The repo example-python will of course be different for each repo.

This call SHOULD be made by the checkout step (see the previous comment issues tagged) and this log.

but we notice that this isn't happening.

A possible solution would be to have the CLI make this call (git config --global --add safe.directory <working_dir>), but not sure if that's the solution we want.

[jonhoo]

I'm seeing the same thing in https://github.com/jonhoo/flurry/actions/runs/7765831951/job/21180957330, and I suspect the problem is related to the job running in a container. I have upgraded a bunch of other repos to v4 without issue (beyond having to set the CODECOV_TOKEN secret time and time again), but none of those have been using a container.

[marekdedic]

I have the issue too, and also only for jobs running in a container :)

[valebes]

I'm having this problem too.

[thomasrockhu-codecov]

Currenlty working on a fix

[thomasrockhu-codecov]

Hi all, I've figured out the root cause, but I don't have a forever solution. For now, a workaround is to add

git config --global --add safe.directory "$GITHUB_WORKSPACE"   # or similar directory

before running Codecov. I didn't expect this to be a problem since the checkout step from GitHub does run this, and you can see some issues actions/runner-images#6775 and actions/checkout#915

I'm looking to add this into the CLI, but I'm not yet 100% we will want this behavior there.

[thomasrockhu-codecov]

The root cause of this error is that all git calls made by the CLI here are failing. The error message is

fatal: detected dubious ownership in repository at '/__w/example-python/example-python'\nTo add an exception for this directory, call:

    git config --global --add safe.directory /__w/example-python/example-python

The repo example-python will of course be different for each repo.

This call SHOULD be made by the checkout step (see the previous comment issues tagged) and this log.

but we notice that this isn't happening.

A possible solution would be to have the CLI make this call (git config --global --add safe.directory <working_dir>), but not sure if that's the solution we want.

[drazisil-codecov]

The --git-service flag is what is needed to fix this issue.

https://github.com/codecov/codecov-cli/blob/da50de0d1dc9f33d01c616dcafd76a0619a0af3c/README.md#create-commit

[dfed]

Workaround worked. Thanks!

[marekdedic]

Hi,
sorry, I am a little bit confused - is this issue actually solved? If there is merely a workaround, I don't think it should be closed...

[thomasrockhu-codecov]

@drazisil-codecov sorry, I combined a few issues into this one here, because the root cause is still this #263 (comment)

[bmulholland]

FYI I'm getting this for runs that don't use a container, but checks out code to a subdirectory instead. Specifically, uses actions/checkout@v4 with path set.

[dcpesses]

In case this helps anyone, it seems codecov-actions@v4 requires the presence of a checked out repo within the same job in order to autodetect the proper values it needs to send when uploading.

In my workflow, the coverage get generated in a "Build & Test" job, which is then handed off to a separate "Coverage Analysis" job as an artifact (using actions/download-artifact@v4).

That setup had worked fine for v3, but when I upgraded to v4, I kept receiving the same "Service not found" error, even after applying the workaround discussed above. Once I added actions/checkout@v4 to the top of the Coverage Analysis job, it worked immediately.

Many thanks to this comment for the inspiration.

[thomasrockhu-codecov]

@dfed it looks like the latest run you made is working, is that accurate?

@jonhoo thanks for sharing your repo, I have tested a fix out on a fork and it looks promising. The fix should be shipped sometime next week.

If you would like to see if this fix will work for you, please use this version of the action

codecov/codecov-action@036b81c4b4e8314b391265bc226440f9e2117e2a

[dfed]

@thomasrockhu-codecov I am unblocked because I used the git config workaround.

I just put up a PR to test your new codecov sha and we'll know if it worked in ~6 minutes

[thomasrockhu-codecov]

@dfed ah! I totally missed that in the config, you're right

[dfed]

@thomasrockhu-codecov good news! codecov/codecov-action@036b81c4b4e8314b391265bc226440f9e2117e2a worked without the workaround!

[marekdedic]

Hi, I don't know if it was supposed to fix it, but I can report that version 4.0.2 does not fix the issue for me - see skaut/shared-drive-mover#1518

[kdarkhan]

@thomasrockhu-codecov can you take a look at this PR codecov/codecov-cli#376?

I thought that it might resolve the issue and it has not been reviewed for 3 weeks already.

[thomasrockhu-codecov]

@marekdedic 4.0.2 does not include the fix mentioned above. 4.0.3 should

[thomasrockhu-codecov]

@kdarkhan I don't believe that fixes the underlying cause, and will break for github_enterprise. This PR should fix the underlying issue for the action.

[kdarkhan]

Got it, thanks for taking a look. I am ok as long as the issue is fixed. I did notice that in the PR you mentioning that the issue will fix it for containers. In my case, I am not using any containers and safe.directory workaround mentioned above does not work for me. The issue I am facing is the one reported in codecov/codecov-action#1287

I am hoping that your PR will fix both issues.

[thomasrockhu-codecov]

@kdarkhan that issue is on my list, thanks for bringing it up. Any more information you can bring to it would be helpful when tackling this week.

[thomasrockhu-codecov]

Ok, a fix has been made in 4.1.0. @dfed you should be able to switch back to v4 with this fix in. fyi @marekdedic

[dfed]

Confirmed this is working for me with v4 and no workaround!

[marekdedic]

Confirming v4.1.0 fixes the issue for me as well, thanks! :)

[thomasrockhu-codecov]

@kdarkhan can you take a look and see if 4.1.0 fixes your issue?

[kdarkhan]

The issue with the "service not found" is fixed but I am still facing an error with the upload file not being found. I think this is specific to running the coverage report in specific subdirectory, so let me move that conversation into codecov/codecov-action#1287

[bmulholland]

@kdarkhan I had to add with: -> working_directory: "path" to codecov's action

[thomasrockhu-codecov]

Closing this out as the original issue has been resolved.