bug: `--container-daemon-socket` parameter does not apply to runner

fiocobbs commented

2025-05-26 00:38:41 +00:00

Can you reproduce the bug on the Forgejo test instance?

N/A

Description

Passing --container-daemon-socket into the ./forgejo-runner executable doesn't seem to affect the utilized socket. It will instead default to unix:///var/run/docker.sock, and since I'm using podman with the socket path being unix://$XDG_RUNTIME_DIR/podman/podman.sock, it will fail. nektos/act does not have this issue. See logs for the command used to test.

I've managed to make a working build for myself, however, I am very not confident it should be merged as-is, for the following reasons:

I'm completely oblivious with the environment variable handling system, and I'm worried I may have caused a regression with my code
I don't know if tracking the extra file(s) is a good idea with a separate repo, and not sure if it'd just be better to inline the function within exec.go

My setup is currently broken as of editing, after I've removed the --container-daemon-socket and tried adding it back in. ~~Looks like the env vars are mutable?~~

I genuinely have no idea how it happened, but a /var/lib/docker.sock was created as a folder, which would not be a valid host. Not sure if the following code patch for docker_socket.go would be warranted to add, but maybe?

// returns socket URI or false if not found any
func socketLocation() (string, bool) {
	if dockerHost, exists := os.LookupEnv("DOCKER_HOST"); exists {
		log.Infof("oijewfoiwfj")
		return dockerHost, true
	}

	for _, p := range CommonSocketLocations {
		if file, err := os.Lstat(os.ExpandEnv(p)); err == nil {
			// START NEW CODE
			if file.IsDir() {
				log.Warnf("socket \"%s\" doesn't seem like a socket, continuing scan...", p)
				continue
			}
			// END NEW CODE

			if strings.HasPrefix(p, `\\.\`) {
				return "npipe://" + filepath.ToSlash(os.ExpandEnv(p)), true
			}
			log.Infof("p %s", p)
			log.Infof("filepath %s", filepath.ToSlash(os.ExpandEnv(p)))
			return "unix://" + filepath.ToSlash(os.ExpandEnv(p)), true
		}
	}

	return "", false
}

If I could get clarification though on what should be done though, I'd be happy to create the two PRs.

The patch relies on copying the docker_socket.go from upstream into the containers repo, and copies some code from nektos/act/root.go.

runner repository Diff

diff --git a/internal/app/cmd/exec.go b/internal/app/cmd/exec.go
index 404f4ae..c06f738 100644
--- a/internal/app/cmd/exec.go
+++ b/internal/app/cmd/exec.go
@@ -13,11 +13,12 @@ import (
 	"strings"
 	"time"
 
-	"github.com/docker/docker/api/types/container"
+	docker_container "github.com/docker/docker/api/types/container"
 	"github.com/joho/godotenv"
 	"github.com/nektos/act/pkg/artifactcache"
 	"github.com/nektos/act/pkg/artifacts"
 	"github.com/nektos/act/pkg/common"
+	"github.com/nektos/act/pkg/container"
 	"github.com/nektos/act/pkg/model"
 	"github.com/nektos/act/pkg/runner"
 	log "github.com/sirupsen/logrus"
@@ -377,6 +378,15 @@ func runExec(ctx context.Context, execArgs *executeArgs) func(cmd *cobra.Command
 			execArgs.artifactServerPath = tempDir
 		}
 
+		// Potentially-uninformed hack to get DOCKER_HOST to work
+		if ret, err := container.GetSocketAndHost(execArgs.containerDaemonSocket); err != nil {
+			log.Warnf("Couldn't get a valid docker connection: %+v", err)
+		} else {
+			os.Setenv("DOCKER_HOST", ret.Host)
+			execArgs.containerDaemonSocket = ret.Socket
+			log.Infof("Using docker host '%s', and daemon socket '%s'", ret.Host, ret.Socket)
+		}
+
 		// run the plan
 		config := &runner.Config{
 			Workdir:               execArgs.Workdir(),
@@ -407,7 +417,7 @@ func runExec(ctx context.Context, execArgs *executeArgs) func(cmd *cobra.Command
 			// EventJSON:                  string(eventJSON),
 			ContainerNamePrefix:        fmt.Sprintf("FORGEJO-ACTIONS-TASK-%s", eventName),
 			ContainerMaxLifetime:       maxLifetime,
-			ContainerNetworkMode:       container.NetworkMode(execArgs.network),
+			ContainerNetworkMode:       docker_container.NetworkMode(execArgs.network),
 			ContainerNetworkEnableIPv6: execArgs.enableIPv6,
 			DefaultActionInstance:      execArgs.defaultActionsURL,
 			PlatformPicker: func(_ []string) string {

I also believe passing in a DOCKER_HOST via CLI doesn't do anything, but this is worth sanity-checking.

Forgejo Version

N/A

Runner Version

6245f71556

How are you running Forgejo?

Attempting to use forgejo-runner as a standalone CLI, to test before pushing to a Forgejo instance

How are you running the Runner?

Compiled from source, at the commit hash defined.

NixOS 25.05 (Warbler) x86_64
podman version 5.4.1

Logs

./forgejo-runner exec --container-daemon-socket unix://$XDG_RUNTIME_DIR/podman/podman.sock

INFO[0000] Using default workflow event: push           
INFO[0000] Planning jobs for event: push                
INFO[0000] cache handler listens on: http://10.8.0.7:46267 
INFO[0000] Start server on http://10.8.0.7:34567        
[build.yaml/Build] [DEBUG] evaluating expression 'success()'
[build.yaml/Build] [DEBUG] expression 'success()' evaluated to 'true'
[build.yaml/Build] 🚀  Start image=debian:12
INFO[0000] Parallel tasks (0) below minimum, setting to 1 
[build.yaml/Build]   🐳  docker pull image=debian:12 platform= username= forcePull=false
[build.yaml/Build] [DEBUG]   🐳  docker pull debian:12
[build.yaml/Build] [DEBUG] Image exists? false
Error: unable to determine if image already exists for image 'debian:12' (): Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?

Workflow file

No response

### Can you reproduce the bug on the Forgejo test instance? *N/A* ### Description Passing `--container-daemon-socket` into the `./forgejo-runner` executable doesn't seem to affect the utilized socket. It will instead default to `unix:///var/run/docker.sock`, and since I'm using podman with the socket path being `unix://$XDG_RUNTIME_DIR/podman/podman.sock`, it will fail. [nektos/act](https://github.com/nektos/act) does not have this issue. See logs for the command used to test. I've managed to make a working build for myself, however, I am very not confident it should be merged as-is, for the following reasons: - I'm completely oblivious with the environment variable handling system, and I'm worried I may have caused a regression with my code - I don't know if tracking the extra file(s) is a good idea with a separate repo, and not sure if it'd just be better to inline the function within `exec.go` - My setup is currently broken as of editing, after I've removed the `--container-daemon-socket` and tried adding it back in. ~~Looks like the env vars are mutable?~~ - I genuinely have no idea how it happened, but a `/var/lib/docker.sock` was created as a folder, which would not be a valid host. Not sure if the following code patch for `docker_socket.go` would be warranted to add, but maybe? <details> ```go // returns socket URI or false if not found any func socketLocation() (string, bool) { if dockerHost, exists := os.LookupEnv("DOCKER_HOST"); exists { log.Infof("oijewfoiwfj") return dockerHost, true } for _, p := range CommonSocketLocations { if file, err := os.Lstat(os.ExpandEnv(p)); err == nil { // START NEW CODE if file.IsDir() { log.Warnf("socket \"%s\" doesn't seem like a socket, continuing scan...", p) continue } // END NEW CODE if strings.HasPrefix(p, `\\.\`) { return "npipe://" + filepath.ToSlash(os.ExpandEnv(p)), true } log.Infof("p %s", p) log.Infof("filepath %s", filepath.ToSlash(os.ExpandEnv(p))) return "unix://" + filepath.ToSlash(os.ExpandEnv(p)), true } } return "", false } ``` </details> If I could get clarification though on what should be done though, I'd be happy to create the two PRs. The patch relies on copying the [docker_socket.go](https://github.com/nektos/act/blob/master/pkg/container/docker_socket.go) from upstream into the containers repo, and copies some code from [nektos/act/root.go](https://github.com/nektos/act/blob/a8ac5293447a2412228f45cf3a430dcd1e92951f/cmd/root.go#L406C1-L412C4). <details><summary>runner repository Diff</summary> ```diff diff --git a/internal/app/cmd/exec.go b/internal/app/cmd/exec.go index 404f4ae..c06f738 100644 --- a/internal/app/cmd/exec.go +++ b/internal/app/cmd/exec.go @@ -13,11 +13,12 @@ import ( "strings" "time" - "github.com/docker/docker/api/types/container" + docker_container "github.com/docker/docker/api/types/container" "github.com/joho/godotenv" "github.com/nektos/act/pkg/artifactcache" "github.com/nektos/act/pkg/artifacts" "github.com/nektos/act/pkg/common" + "github.com/nektos/act/pkg/container" "github.com/nektos/act/pkg/model" "github.com/nektos/act/pkg/runner" log "github.com/sirupsen/logrus" @@ -377,6 +378,15 @@ func runExec(ctx context.Context, execArgs *executeArgs) func(cmd *cobra.Command execArgs.artifactServerPath = tempDir } + // Potentially-uninformed hack to get DOCKER_HOST to work + if ret, err := container.GetSocketAndHost(execArgs.containerDaemonSocket); err != nil { + log.Warnf("Couldn't get a valid docker connection: %+v", err) + } else { + os.Setenv("DOCKER_HOST", ret.Host) + execArgs.containerDaemonSocket = ret.Socket + log.Infof("Using docker host '%s', and daemon socket '%s'", ret.Host, ret.Socket) + } + // run the plan config := &runner.Config{ Workdir: execArgs.Workdir(), @@ -407,7 +417,7 @@ func runExec(ctx context.Context, execArgs *executeArgs) func(cmd *cobra.Command // EventJSON: string(eventJSON), ContainerNamePrefix: fmt.Sprintf("FORGEJO-ACTIONS-TASK-%s", eventName), ContainerMaxLifetime: maxLifetime, - ContainerNetworkMode: container.NetworkMode(execArgs.network), + ContainerNetworkMode: docker_container.NetworkMode(execArgs.network), ContainerNetworkEnableIPv6: execArgs.enableIPv6, DefaultActionInstance: execArgs.defaultActionsURL, PlatformPicker: func(_ []string) string { ``` </details> I also believe passing in a `DOCKER_HOST` via CLI doesn't do anything, but this is worth sanity-checking. ### Forgejo Version N/A ### Runner Version 6245f71556178342f0163cd88d0a0070f1eabf53 ### How are you running Forgejo? Attempting to use forgejo-runner as a standalone CLI, to test before pushing to a Forgejo instance ### How are you running the Runner? Compiled from source, at the commit hash defined. - NixOS 25.05 (Warbler) x86_64 - podman version 5.4.1 ### Logs `./forgejo-runner exec --container-daemon-socket unix://$XDG_RUNTIME_DIR/podman/podman.sock` ``` INFO[0000] Using default workflow event: push INFO[0000] Planning jobs for event: push INFO[0000] cache handler listens on: http://10.8.0.7:46267 INFO[0000] Start server on http://10.8.0.7:34567 [build.yaml/Build] [DEBUG] evaluating expression 'success()' [build.yaml/Build] [DEBUG] expression 'success()' evaluated to 'true' [build.yaml/Build] 🚀 Start image=debian:12 INFO[0000] Parallel tasks (0) below minimum, setting to 1 [build.yaml/Build] 🐳 docker pull image=debian:12 platform= username= forcePull=false [build.yaml/Build] [DEBUG] 🐳 docker pull debian:12 [build.yaml/Build] [DEBUG] Image exists? false Error: unable to determine if image already exists for image 'debian:12' (): Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running? ``` ### Workflow file _No response_

fiocobbs added the

Kind/Bug

label

2025-05-26 00:38:41 +00:00

fiocobbs changed title from ~~bug: --container-daemon-socket parameter does not apply to runner~~ to bug: --container-daemon-socket parameter does not apply to runner

2025-05-26 00:51:59 +00:00

earl-warren commented

2025-05-26 05:52:45 +00:00

Contributor

Passing --container-daemon-socket into the ./forgejo-runner executable doesn't seem to affect the utilized socket. It will instead default to unix:///var/run/docker.sock,

I would not be surprised if the code handling variables overrides the parameter.

I did not look closely enough to be sure but from a distance it looks like you are on the right track. If you send a PR to fix this I'll take a closer look and this can be fixed on the next release.

Note that there are assorted tests at https://code.forgejo.org/forgejo/end-to-end/src/branch/main/actions (example-with-docker-host etc.) and one could be added to verify the argument is processed as it should. Alternatively the argument could be removed for simplification, unless it is a blocker.

> Passing --container-daemon-socket into the ./forgejo-runner executable doesn't seem to affect the utilized socket. It will instead default to unix:///var/run/docker.sock, I would not be surprised if the code handling variables overrides the parameter. I did not look closely enough to be sure but from a distance it looks like you are on the right track. If you send a PR to fix this I'll take a closer look and this can be fixed on the next release. Note that there are assorted tests at https://code.forgejo.org/forgejo/end-to-end/src/branch/main/actions (example-with-docker-host etc.) and one could be added to verify the argument is processed as it should. Alternatively the argument could be removed for simplification, unless it is a blocker.

el0n commented

2025-05-26 15:54:42 +00:00

@fiocobbs wrote in #577 (comment):

Passing --container-daemon-socket into the ./forgejo-runner executable doesn't seem to affect the utilized socket. It will instead default to unix:///var/run/docker.sock,

Just install the podman-docker package. It will provide a link reference ....

$ cat /usr/lib/tmpfiles.d/podman-docker.conf
L+  %t/docker.sock   -    -    -     -   %t/podman/podman.sock

@fiocobbs wrote in https://code.forgejo.org/forgejo/runner/issues/577#issue-7252: > ### > > Passing `--container-daemon-socket` into the `./forgejo-runner` executable doesn't seem to affect the utilized socket. It will instead default to `unix:///var/run/docker.sock`, Just install the podman-docker package. It will provide a link reference .... ``` $ cat /usr/lib/tmpfiles.d/podman-docker.conf L+ %t/docker.sock - - - - %t/podman/podman.sock ```

fiocobbs referenced this issue from a pull request that will close it,

2025-05-27 00:19:52 +00:00

fix: apply --container-daemon-socket correctly [skip cascade] #581

diem commented

2025-06-23 09:35:53 +00:00

@el0n wrote in #577 (comment):

@fiocobbs wrote in #577 (comment):

Passing --container-daemon-socket into the ./forgejo-runner executable doesn't seem to affect the utilized socket. It will instead default to unix:///var/run/docker.sock,

Just install the podman-docker package. It will provide a link reference ....
$ cat /usr/lib/tmpfiles.d/podman-docker.conf
L+  %t/docker.sock   -    -    -     -   %t/podman/podman.sock

Unfortunately this doesn't "fix" the issue since podman can run in both "rootfull" and "rootless" mode. The latter the socket is located at unix://$XDG_RUNTIME_DIR/podman/podman.sock , which for example on my machine translates to unix:///run/user/1000/podman/podman.sock while /var/run/docker.sock is a symlink to /run/podman/podman.sock and /run/podman have 0700 permission set which the user have no access, unless you run sudo forgejo-runner .... which any best practices will say otherwise.

I would rather see a fix to use either DOCKER_HOST or DOCKER_SOCKET env variables

@el0n wrote in https://code.forgejo.org/forgejo/runner/issues/577#issuecomment-41034: > @fiocobbs wrote in #577 (comment): > > > ### [](#heading) > > Passing `--container-daemon-socket` into the `./forgejo-runner` executable doesn't seem to affect the utilized socket. It will instead default to `unix:///var/run/docker.sock`, > > Just install the podman-docker package. It will provide a link reference .... > > ```text > $ cat /usr/lib/tmpfiles.d/podman-docker.conf > L+ %t/docker.sock - - - - %t/podman/podman.sock > ``` Unfortunately this doesn't "fix" the issue since podman can run in both "rootfull" and "rootless" mode. The latter the socket is located at `unix://$XDG_RUNTIME_DIR/podman/podman.sock` , which for example on my machine translates to `unix:///run/user/1000/podman/podman.sock` while `/var/run/docker.sock` is a symlink to `/run/podman/podman.sock` and `/run/podman` have `0700` permission set which the user have no access, unless you run `sudo forgejo-runner ....` which any best practices will say otherwise. I would rather see a fix to use either DOCKER_HOST or DOCKER_SOCKET env variables

el0n commented

2025-06-23 20:56:28 +00:00

What matters is how it looks like in the container. Here a snippet of a user quadlet file:
Volume=%t/podman/podman.sock:/var/run/docker.sock

What matters is how it looks like in the container. Here a snippet of a user quadlet file: `Volume=%t/podman/podman.sock:/var/run/docker.sock`

earl-warren added the

Reviewed

Confirmed

label

2025-07-02 10:49:58 +00:00

earl-warren added the

Priority

Medium

label

2025-07-11 12:34:20 +00:00

earl-warren referenced this issue from a commit

2025-07-28 17:10:27 +00:00

fix: use new platforms after survey, check working dir for `.actrc` (#577)

earl-warren commented

2025-08-16 10:22:08 +00:00

Contributor

@fiocobbs @c8h4 there are two pull requests in the same area at the moment:

#598
#581

They were rebased against the latest main branch today and pass tests (no big surprise 😁). I'll comment here since the review applies to both.

The logic of how the container server is determined in Forgejo runner is convoluted. And it turns out the --container-daemon-socket does not do anything, because it is always overridden by the existing logic.

Showcase

on: [push]
jobs:
  test:
    runs-on: docker
    steps:
      - run: echo All good!

earl-warren:/tmp/g$ forgejo-runner-9.0.3 exec --workflows /tmp/a.yml
INFO[0000] Using the only detected workflow event: push 
INFO[0000] Planning jobs for event: push                
INFO[0000] cache handler listens on: http://192.168.3.6:39645 
[a.yml/test] 🚀  Start image=node:20-bullseye
[a.yml/test]   🐳  docker pull image=node:20-bullseye platform= username= forcePull=false
[a.yml/test] Cleaning up network for job test, and network name is: FORGEJO-ACTIONS-TASK-push_WORKFLOW-40eb3fd3849866be6240d5d99e801ba8be6003ba7e2f976b44388af58dedd5a0_JOB-test-test-network
[a.yml/test]   🐳  docker create image=node:20-bullseye platform= entrypoint=["tail" "-f" "/dev/null"] cmd=[] network="FORGEJO-ACTIONS-TASK-push_WORKFLOW-40eb3fd3849866be6240d5d99e801ba8be6003ba7e2f976b44388af58dedd5a0_JOB-test-test-network"
[a.yml/test]   🐳  docker run image=node:20-bullseye platform= entrypoint=["tail" "-f" "/dev/null"] cmd=[] network="FORGEJO-ACTIONS-TASK-push_WORKFLOW-40eb3fd3849866be6240d5d99e801ba8be6003ba7e2f976b44388af58dedd5a0_JOB-test-test-network"
[a.yml/test] ⭐ Run Main echo All good!
[a.yml/test]   🐳  docker exec cmd=[sh /var/run/act/b072be7d355cfb3b.sh] user= workdir=
[a.yml/test]   🐳  docker exec cmd=[bash --noprofile --norc -e -o pipefail /var/run/act/workflow/0.sh] user= workdir=
| All good!
[a.yml/test]   ✅  Success - Main echo All good!
[a.yml/test] 🏁  Job succeeded
INFO[0001] Cleaning up network for job test, and network name is: FORGEJO-ACTIONS-TASK-push_WORKFLOW-40eb3fd3849866be6240d5d99e801ba8be6003ba7e2f976b44388af58dedd5a0_JOB-test-test-network

It still works fine with forgejo-runner-9.0.3 exec --container-daemon-socket=/aaa --workflows /tmp/a.yml. But it breaks with:

earl-warren:/tmp/g$ DOCKER_HOST=/aaaa forgejo-runner-9.0.3 exec --workflows /tmp/a.yml
INFO[0000] Using the only detected workflow event: push 
INFO[0000] Planning jobs for event: push                
INFO[0000] cache handler listens on: http://192.168.3.6:35655 
[a.yml/test] 🚀  Start image=node:20-bullseye
[a.yml/test]   🐳  docker pull image=node:20-bullseye platform= username= forcePull=false
Error: unable to determine if image already exists for image 'node:20-bullseye' (): failed to connect to docker daemon: unable to parse docker host `/aaaa`

Instead of trying to fix it and since there is already a way to specify the desired value from the environment, I think the option should just be removed. The variable DOCKER_HOST is documented in a specific case. But it may be good to add a line about it in the help of forgejo-runner exec.

What do you think?

@fiocobbs @c8h4 there are two pull requests in the same area at the moment: - https://code.forgejo.org/forgejo/runner/pulls/598 - https://code.forgejo.org/forgejo/runner/pulls/581 They were rebased against the latest main branch today and pass tests (no big surprise 😁). I'll comment here since the review applies to both. The logic of how the container server is determined in Forgejo runner is convoluted. And it turns out the `--container-daemon-socket` does not do anything, because it is always overridden by the existing logic. ### Showcase ```yaml on: [push] jobs: test: runs-on: docker steps: - run: echo All good! ``` ```sh earl-warren:/tmp/g$ forgejo-runner-9.0.3 exec --workflows /tmp/a.yml INFO[0000] Using the only detected workflow event: push INFO[0000] Planning jobs for event: push INFO[0000] cache handler listens on: http://192.168.3.6:39645 [a.yml/test] 🚀 Start image=node:20-bullseye [a.yml/test] 🐳 docker pull image=node:20-bullseye platform= username= forcePull=false [a.yml/test] Cleaning up network for job test, and network name is: FORGEJO-ACTIONS-TASK-push_WORKFLOW-40eb3fd3849866be6240d5d99e801ba8be6003ba7e2f976b44388af58dedd5a0_JOB-test-test-network [a.yml/test] 🐳 docker create image=node:20-bullseye platform= entrypoint=["tail" "-f" "/dev/null"] cmd=[] network="FORGEJO-ACTIONS-TASK-push_WORKFLOW-40eb3fd3849866be6240d5d99e801ba8be6003ba7e2f976b44388af58dedd5a0_JOB-test-test-network" [a.yml/test] 🐳 docker run image=node:20-bullseye platform= entrypoint=["tail" "-f" "/dev/null"] cmd=[] network="FORGEJO-ACTIONS-TASK-push_WORKFLOW-40eb3fd3849866be6240d5d99e801ba8be6003ba7e2f976b44388af58dedd5a0_JOB-test-test-network" [a.yml/test] ⭐ Run Main echo All good! [a.yml/test] 🐳 docker exec cmd=[sh /var/run/act/b072be7d355cfb3b.sh] user= workdir= [a.yml/test] 🐳 docker exec cmd=[bash --noprofile --norc -e -o pipefail /var/run/act/workflow/0.sh] user= workdir= | All good! [a.yml/test] ✅ Success - Main echo All good! [a.yml/test] 🏁 Job succeeded INFO[0001] Cleaning up network for job test, and network name is: FORGEJO-ACTIONS-TASK-push_WORKFLOW-40eb3fd3849866be6240d5d99e801ba8be6003ba7e2f976b44388af58dedd5a0_JOB-test-test-network ``` It still works fine with `forgejo-runner-9.0.3 exec --container-daemon-socket=/aaa --workflows /tmp/a.yml`. But it breaks with: ```sh earl-warren:/tmp/g$ DOCKER_HOST=/aaaa forgejo-runner-9.0.3 exec --workflows /tmp/a.yml INFO[0000] Using the only detected workflow event: push INFO[0000] Planning jobs for event: push INFO[0000] cache handler listens on: http://192.168.3.6:35655 [a.yml/test] 🚀 Start image=node:20-bullseye [a.yml/test] 🐳 docker pull image=node:20-bullseye platform= username= forcePull=false Error: unable to determine if image already exists for image 'node:20-bullseye' (): failed to connect to docker daemon: unable to parse docker host `/aaaa` ``` Instead of trying to fix it and since there is already a way to specify the desired value from the environment, I think the option should just be removed. The variable `DOCKER_HOST` is [documented](https://forgejo.org/docs/next/admin/actions/runner-installation/#setting-up-the-container-environment) in a specific case. But it may be good to add a line about it in the help of `forgejo-runner exec`. What do you think?

earl-warren referenced this issue

2025-08-16 10:36:01 +00:00

cmd: exec: prefer docker socket path from DOCKER_HOST if specified #598

earl-warren referenced this issue

2025-08-16 10:36:12 +00:00

fix: apply --container-daemon-socket correctly [skip cascade] #581

earl-warren commented

2025-08-29 14:00:30 +00:00

Contributor

@fiocobbs @c8h4 if you would like me to try and fix this please let me know. Otherwise it can wait until you have time. 🙏

earl-warren commented

2025-09-16 21:07:36 +00:00

Contributor

Instead of trying to fix it and since there is already a way to specify the desired value from the environment, I think the option should just be removed. The variable DOCKER_HOST is documented in a specific case. But it may be good to add a line about it in the help of forgejo-runner exec.

I went ahead and did that at #1004

> Instead of trying to fix it and since there is already a way to specify the desired value from the environment, I think the option should just be removed. The variable DOCKER_HOST is documented in a specific case. But it may be good to add a line about it in the help of forgejo-runner exec. I went ahead and did that at https://code.forgejo.org/forgejo/runner/pulls/1004

❤️ 1

earl-warren referenced this issue

2025-09-16 21:08:00 +00:00

chore: update forgejo-runner exec defaults and reword the help #1004

earl-warren commented

2025-09-18 08:21:18 +00:00

Contributor

The scope of the proposed pull request was too wide, I'll submit another.

earl-warren referenced this issue from a commit

2025-09-18 08:31:09 +00:00

chore: document that DOCKER_HOST should be used instead of --container-daemon-socket

earl-warren referenced this issue from a pull request that will close it,