fix(git-sync): run auto-pull as the admin who enabled it (#11121)

* fix(git-sync): run auto-pull as the admin who enabled it

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* fix(git-sync): audit the admin grant fork pulls make

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* chore: bump ee ref for the post-commit fork grant audit

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* fix(git-sync): address review nits on the auto-pull stamp

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

* chore: update ee-repo-ref to ccada062c072d7b74894b63863728fd1ef9bdffd

This commit updates the EE repository reference after PR #799 was merged in windmill-ee-private.

Previous ee-repo-ref: 7cee30f0cf12721cba551cd754dc817444810470

New ee-repo-ref: ccada062c072d7b74894b63863728fd1ef9bdffd

Automated by sync-ee-ref workflow.

---------

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Co-authored-by: windmill-internal-app[bot] <windmill-internal-app[bot]@users.noreply.github.com>
This commit is contained in:
hugocasa
2026-09-15 10:34:57 +02:00
committed by GitHub
co-authored by Claude Opus 5 windmill-internal-app[bot]
parent 082d897328
commit 69e6efd875
20 changed files with 481 additions and 43 deletions
@@ -1,6 +1,6 @@
{
"db_name": "PostgreSQL",
"query": "SELECT username, email FROM usr WHERE workspace_id = $1 AND is_admin = true AND operator = false AND disabled = false ORDER BY username LIMIT 1",
"query": "SELECT u.username, u.email FROM usr u WHERE u.workspace_id = $1 AND u.is_admin AND NOT u.operator AND NOT u.disabled AND NOT EXISTS (SELECT 1 FROM password p WHERE p.email = u.email AND p.disabled) ORDER BY u.username LIMIT 1",
"describe": {
"columns": [
{
@@ -24,5 +24,5 @@
false
]
},
"hash": "3202bed875693ae923f496272cd8ad89b2f17a9d3ef4659c2d2284415177b32c"
"hash": "0c4dc0e9dc159fac7e41492c78a4e4e0b12b105d4475d7eba2d3a9573b93e388"
}
@@ -1,26 +0,0 @@
{
"db_name": "PostgreSQL",
"query": "SELECT COALESCE(username, split_part(email, '@', 1)) AS \"username!\", email FROM password WHERE super_admin = true AND disabled = false ORDER BY email LIMIT 1",
"describe": {
"columns": [
{
"ordinal": 0,
"name": "username!",
"type_info": "Varchar"
},
{
"ordinal": 1,
"name": "email",
"type_info": "Varchar"
}
],
"parameters": {
"Left": []
},
"nullable": [
null,
false
]
},
"hash": "17cdf02b4912078459526205849246fd2bdf9e3fce89852120aa4ad4ad6abfc3"
}
@@ -0,0 +1,22 @@
{
"db_name": "PostgreSQL",
"query": "SELECT super_admin AS \"super_admin!\" FROM password WHERE email = $1 AND disabled = false",
"describe": {
"columns": [
{
"ordinal": 0,
"name": "super_admin!",
"type_info": "Bool"
}
],
"parameters": {
"Left": [
"Text"
]
},
"nullable": [
false
]
},
"hash": "3fe41e2a72d02613a2b1c1c44fb0a7b681d101d286adfc6ff4548d1b9fdbab8c"
}
@@ -0,0 +1,23 @@
{
"db_name": "PostgreSQL",
"query": "SELECT r->'auto_pull'->>'enabled_by' AS \"enabled_by\"\n FROM workspace_settings, jsonb_array_elements(git_sync->'repositories') r\n WHERE workspace_id = $1 AND r->>'git_repo_resource_path' = $2",
"describe": {
"columns": [
{
"ordinal": 0,
"name": "enabled_by",
"type_info": "Text"
}
],
"parameters": {
"Left": [
"Text",
"Text"
]
},
"nullable": [
null
]
},
"hash": "6dc8032100a28c4a6e843370038dfec43dc8a042f9d95e63bf84b2dc15a72165"
}
@@ -0,0 +1,47 @@
{
"db_name": "PostgreSQL",
"query": "SELECT u.username, u.is_admin, u.operator, u.disabled, EXISTS (SELECT 1 FROM password p WHERE p.email = u.email AND p.disabled) AS \"instance_disabled!\" FROM usr u WHERE u.workspace_id = $1 AND u.email = $2",
"describe": {
"columns": [
{
"ordinal": 0,
"name": "username",
"type_info": "Varchar"
},
{
"ordinal": 1,
"name": "is_admin",
"type_info": "Bool"
},
{
"ordinal": 2,
"name": "operator",
"type_info": "Bool"
},
{
"ordinal": 3,
"name": "disabled",
"type_info": "Bool"
},
{
"ordinal": 4,
"name": "instance_disabled!",
"type_info": "Bool"
}
],
"parameters": {
"Left": [
"Text",
"Text"
]
},
"nullable": [
false,
false,
false,
false,
null
]
},
"hash": "8bdfc02e7be54c2b610fed11cce75f8d2f1665d329b8adea424b3a6f1e5c7013"
}
@@ -0,0 +1,24 @@
{
"db_name": "PostgreSQL",
"query": "INSERT INTO usr (workspace_id, username, email, is_admin, operator)\n SELECT $1::varchar, u.username, u.email, true, false FROM usr u\n WHERE u.workspace_id = $2 AND u.email = $3 AND u.is_admin AND NOT u.operator AND NOT u.disabled\n AND NOT EXISTS (SELECT 1 FROM password p WHERE p.email = u.email AND p.disabled)\n AND NOT EXISTS (SELECT 1 FROM usr f WHERE f.workspace_id = $1::varchar AND f.email = $3)\n ON CONFLICT DO NOTHING\n RETURNING username",
"describe": {
"columns": [
{
"ordinal": 0,
"name": "username",
"type_info": "Varchar"
}
],
"parameters": {
"Left": [
"Varchar",
"Text",
"Text"
]
},
"nullable": [
false
]
},
"hash": "a80a18774baf36d09b07da1e4e30baab26b8160929aea6d8f5f226e6ec4f8bd8"
}
@@ -0,0 +1,23 @@
{
"db_name": "PostgreSQL",
"query": "SELECT EXISTS (SELECT 1 FROM usr WHERE workspace_id = $1 AND username = $2) AS \"claimed!\"",
"describe": {
"columns": [
{
"ordinal": 0,
"name": "claimed!",
"type_info": "Bool"
}
],
"parameters": {
"Left": [
"Text",
"Text"
]
},
"nullable": [
null
]
},
"hash": "b675c20bb7a15bec5e9a34d7ddf347e22196e46727487f35634af079c71c2bef"
}
+1
View File
@@ -15727,6 +15727,7 @@ dependencies = [
"tokio",
"tracing",
"uuid",
"windmill-audit",
"windmill-common",
"windmill-dep-map",
"windmill-queue",
+1 -1
View File
@@ -1 +1 @@
0780955effb657807d14f0eb503cba1d49cee007
ccada062c072d7b74894b63863728fd1ef9bdffd
+46
View File
@@ -0,0 +1,46 @@
-- A parent workspace whose auto-pulled repository was last saved by alice, and a fork of
-- it holding only carol, the non-admin who created it. aaron is an admin who sorts before
-- alice; bob is no longer an admin; dora is a workspace admin deactivated on the instance.
-- sam and sue are instance superadmins who are not members: sam's instance username is
-- carol's, sue's is unclaimed.
INSERT INTO workspace (id, name, owner) VALUES ('ap-parent', 'ap-parent', 'alice@windmill.dev');
INSERT INTO workspace (id, name, owner, parent_workspace_id)
VALUES ('wm-fork-feat', 'feat', 'carol@windmill.dev', 'ap-parent');
INSERT INTO group_ (workspace_id, name, summary, extra_perms) VALUES
('ap-parent', 'all', 'All users', '{}'),
('wm-fork-feat', 'all', 'All users', '{}');
INSERT INTO password (email, password_hash, login_type, super_admin, verified, name, disabled) VALUES
('aaron@windmill.dev', 'not-a-real-hash', 'password', false, true, 'aaron', false),
('alice@windmill.dev', 'not-a-real-hash', 'password', false, true, 'alice', false),
('bob@windmill.dev', 'not-a-real-hash', 'password', false, true, 'bob', false),
('carol@windmill.dev', 'not-a-real-hash', 'password', false, true, 'carol', false),
('dora@windmill.dev', 'not-a-real-hash', 'password', false, true, 'dora', true);
INSERT INTO password (email, password_hash, login_type, super_admin, verified, name, disabled, username) VALUES
('sam@windmill.dev', 'not-a-real-hash', 'password', true, true, 'sam', false, 'carol'),
('sue@windmill.dev', 'not-a-real-hash', 'password', true, true, 'sue', false, 'sue');
INSERT INTO usr (workspace_id, email, username, is_admin) VALUES
('ap-parent', 'aaron@windmill.dev', 'aaron', true),
('ap-parent', 'alice@windmill.dev', 'alice', true),
('ap-parent', 'bob@windmill.dev', 'bob', false),
('ap-parent', 'carol@windmill.dev', 'carol', false),
('ap-parent', 'dora@windmill.dev', 'dora', true),
('wm-fork-feat', 'carol@windmill.dev', 'carol', false);
INSERT INTO resource (workspace_id, path, value, resource_type, extra_perms, created_by) VALUES
('ap-parent', 'u/alice/repo', '{"url": "https://github.com/test/repo.git", "branch": "main"}',
'git_repository', '{}', 'alice'),
('wm-fork-feat', 'u/alice/repo', '{"url": "https://github.com/test/repo.git", "branch": "main"}',
'git_repository', '{}', 'alice');
INSERT INTO workspace_settings (workspace_id, git_sync) VALUES
('ap-parent', '{"repositories":[{"git_repo_resource_path":"$res:u/alice/repo",
"use_individual_branch":false,"group_by_folder":false,
"auto_pull":{"enabled":true,"mode":"polling","sync_forks":true,
"enabled_by":"alice@windmill.dev"}}]}'),
('wm-fork-feat', '{"repositories":[{"git_repo_resource_path":"$res:u/alice/repo",
"use_individual_branch":false,"group_by_folder":false}]}');
+194
View File
@@ -0,0 +1,194 @@
//! An automatic pull runs as the admin stamped on the repository's settings, never as
//! someone picked from the workspace, and stops once that admin is revoked. A fork's
//! pull runs as the parent's pull identity, added to the fork first.
#![cfg(all(feature = "enterprise", feature = "private"))]
use sqlx::{Pool, Postgres};
use windmill_common::workspaces::GitRepositorySettings;
use windmill_git_sync::{reconcile_and_enqueue_pull, reconcile_fork_branch_pull};
const PARENT: &str = "ap-parent";
const FORK: &str = "wm-fork-feat";
const REPO: &str = "$res:u/alice/repo";
fn repo_enabled_by(email: &str) -> GitRepositorySettings {
serde_json::from_value(serde_json::json!({
"git_repo_resource_path": REPO,
"use_individual_branch": false,
"group_by_folder": false,
"auto_pull": { "enabled": true, "enabled_by": email }
}))
.expect("repository settings")
}
/// `(created_by, permissioned_as, permissioned_as_email)` of every pull job in `w_id`.
async fn pull_identities(
db: &Pool<Postgres>,
w_id: &str,
) -> anyhow::Result<Vec<(String, String, Option<String>)>> {
Ok(sqlx::query_as(
"SELECT created_by, permissioned_as, permissioned_as_email FROM v2_job \
WHERE workspace_id = $1 AND kind = 'deploymentcallback'",
)
.bind(w_id)
.fetch_all(db)
.await?)
}
fn identity(username: &str) -> (String, String, Option<String>) {
(
username.to_string(),
format!("u/{username}"),
Some(format!("{username}@windmill.dev")),
)
}
async fn recorded_pull_error(db: &Pool<Postgres>, w_id: &str) -> anyhow::Result<String> {
let git_sync: serde_json::Value =
sqlx::query_scalar("SELECT git_sync FROM workspace_settings WHERE workspace_id = $1")
.bind(w_id)
.fetch_one(db)
.await?;
Ok(
git_sync["repositories"][0]["auto_pull"]["last_pull_status"]["error"]
.as_str()
.unwrap_or_default()
.to_string(),
)
}
#[sqlx::test(fixtures("git_sync_autopull_identity"))]
async fn pull_runs_as_the_admin_who_enabled_it(db: Pool<Postgres>) -> anyhow::Result<()> {
let job = reconcile_and_enqueue_pull(
&db,
PARENT,
&repo_enabled_by("alice@windmill.dev"),
"main",
"abc123",
None,
)
.await?;
assert!(job.is_some());
assert_eq!(pull_identities(&db, PARENT).await?, vec![identity("alice")]);
Ok(())
}
/// bob was demoted in the workspace; dora is still a workspace admin but deactivated on
/// the instance. Neither may run the pull, and the failure lands on the status rather
/// than as an error, which a webhook delivery would turn into a failed response.
#[sqlx::test(fixtures("git_sync_autopull_identity"))]
async fn pull_fails_on_the_status_once_the_enabling_admin_is_revoked(
db: Pool<Postgres>,
) -> anyhow::Result<()> {
for email in ["bob@windmill.dev", "dora@windmill.dev"] {
let job = reconcile_and_enqueue_pull(
&db,
PARENT,
&repo_enabled_by(email),
"main",
"abc123",
None,
)
.await?;
assert!(job.is_none(), "{email} must not run the pull");
let error = recorded_pull_error(&db, PARENT).await?;
assert!(error.contains(email), "{error}");
}
assert!(pull_identities(&db, PARENT).await?.is_empty());
Ok(())
}
#[sqlx::test(fixtures("git_sync_autopull_identity"))]
async fn fork_pull_runs_as_the_parent_admin_added_to_the_fork(
db: Pool<Postgres>,
) -> anyhow::Result<()> {
let job = reconcile_fork_branch_pull(&db, PARENT, REPO, "wm-fork/main/feat", "main", "abc123")
.await?;
assert!(
job.is_some(),
"the fork branch must route to the fork and enqueue"
);
let (is_admin, in_all): (bool, bool) = sqlx::query_as(
"SELECT u.is_admin, EXISTS (SELECT 1 FROM usr_to_group g \
WHERE g.workspace_id = u.workspace_id AND g.usr = u.username AND g.group_ = 'all') \
FROM usr u WHERE u.workspace_id = $1 AND u.email = 'alice@windmill.dev'",
)
.bind(FORK)
.fetch_one(&db)
.await?;
assert!(
is_admin && in_all,
"alice must be an admin member of the fork"
);
assert_eq!(pull_identities(&db, FORK).await?, vec![identity("alice")]);
let grants: i64 = sqlx::query_scalar(
"SELECT count(*) FROM audit_partitioned WHERE workspace_id = $1 \
AND operation = 'users.git_sync_fork_add' AND resource = 'alice@windmill.dev'",
)
.bind(FORK)
.fetch_one(&db)
.await?;
assert_eq!(grants, 1, "adding alice to the fork must be audited");
Ok(())
}
/// A superadmin who is not a member runs the pull under their instance username, and
/// `u/<username>` resolves through the workspace's members first. sam's instance username
/// is carol's, so sam's stamp must not run the pull as carol; sue's is unclaimed.
#[sqlx::test(fixtures("git_sync_autopull_identity"))]
async fn a_non_member_superadmin_runs_the_pull_only_under_an_unclaimed_username(
db: Pool<Postgres>,
) -> anyhow::Result<()> {
let job = reconcile_and_enqueue_pull(
&db,
PARENT,
&repo_enabled_by("sam@windmill.dev"),
"main",
"abc123",
None,
)
.await?;
assert!(job.is_none(), "sam's username belongs to carol");
assert!(pull_identities(&db, PARENT).await?.is_empty());
let job = reconcile_and_enqueue_pull(
&db,
PARENT,
&repo_enabled_by("sue@windmill.dev"),
"main",
"abc123",
None,
)
.await?;
assert!(job.is_some());
assert_eq!(pull_identities(&db, PARENT).await?, vec![identity("sue")]);
Ok(())
}
/// With no stamp on the parent, the fork pull still runs as the parent's first active
/// admin: the fork holds only its non-admin creator, so no identity resolved in the fork
/// could run it.
#[sqlx::test(fixtures("git_sync_autopull_identity"))]
async fn unstamped_fork_pull_runs_as_the_parents_first_admin(
db: Pool<Postgres>,
) -> anyhow::Result<()> {
sqlx::query(
"UPDATE workspace_settings SET git_sync = git_sync #- '{repositories,0,auto_pull,enabled_by}' \
WHERE workspace_id = $1",
)
.bind(PARENT)
.execute(&db)
.await?;
let job = reconcile_fork_branch_pull(&db, PARENT, REPO, "wm-fork/main/feat", "main", "abc123")
.await?;
assert!(
job.is_some(),
"an unstamped parent must still sync its forks"
);
assert_eq!(pull_identities(&db, FORK).await?, vec![identity("aaron")]);
Ok(())
}
@@ -923,6 +923,7 @@ async fn test_pull_stays_on_the_workspace_lane(db: Pool<Postgres>) -> anyhow::Re
&db,
"test-workspace",
&repo,
("test-user", "test@windmill.dev"),
None,
false,
None,
@@ -910,12 +910,15 @@ fn redact_git_sync_webhook_secrets(git_sync: &mut serde_json::Value) {
}
/// Zero the server-owned auto-pull fields (webhook id/secret/url/error, synced
/// sha, last pull status) on a client-supplied `AutoPullSettings`. The client only
/// controls `enabled` / `mode` / `poll_interval_s`; the rest is written by the
/// server (webhook creation, poller) and must never be trusted from the request —
/// otherwise a caller could inject a webhook id/secret or fake sync state.
fn clear_client_supplied_auto_pull_state(
/// sha, last pull status) on a client-supplied `AutoPullSettings`, and stamp who
/// pulls run as: `saver_email` while auto pull is on. The client only controls
/// `enabled` / `mode` / `poll_interval_s`; the rest is written by the server (webhook
/// creation, poller, this save) and must never be trusted from the request —
/// otherwise a caller could inject a webhook id/secret, fake sync state, or pick who
/// pulls run as.
fn sanitize_client_auto_pull(
auto_pull: &mut windmill_common::workspaces::AutoPullSettings,
saver_email: &str,
) {
auto_pull.webhook_id = None;
auto_pull.webhook_secret = None;
@@ -923,6 +926,28 @@ fn clear_client_supplied_auto_pull_state(
auto_pull.webhook_error = None;
auto_pull.last_synced_sha = std::collections::HashMap::new();
auto_pull.last_pull_status = None;
auto_pull.enabled_by = auto_pull.enabled.then(|| saver_email.to_string());
}
#[cfg(test)]
mod sanitize_client_auto_pull_tests {
use windmill_common::workspaces::AutoPullSettings;
#[test]
fn a_save_stamps_the_saver_over_any_client_supplied_stamp() {
let mut ap = AutoPullSettings {
enabled: true,
enabled_by: Some("forged@example.com".to_string()),
..Default::default()
};
super::sanitize_client_auto_pull(&mut ap, "saver@example.com");
assert_eq!(ap.enabled_by.as_deref(), Some("saver@example.com"));
ap.enabled = false;
ap.enabled_by = Some("forged@example.com".to_string());
super::sanitize_client_auto_pull(&mut ap, "saver@example.com");
assert_eq!(ap.enabled_by, None, "auto pull off carries no stamp");
}
}
/// Whether a git-sync repository tracking `tracked` rules out `label_branch` as a dev workspace's
@@ -3983,7 +4008,7 @@ async fn edit_git_sync_config(
// stay clean.
for repo in git_sync_settings.repositories.iter_mut() {
if let Some(ap) = repo.auto_pull.as_mut() {
clear_client_supplied_auto_pull_state(ap);
sanitize_client_auto_pull(ap, &authed.email);
}
repo.open_pr_error = None;
repo.credential = None;
@@ -4230,7 +4255,7 @@ async fn edit_git_sync_repository(
// existing repo re-derives it from the DB (carried over below) and a new one
// starts clean.
if let Some(ap) = new_config.repository.auto_pull.as_mut() {
clear_client_supplied_auto_pull_state(ap);
sanitize_client_auto_pull(ap, &authed.email);
}
new_config.repository.open_pr_error = None;
new_config.repository.credential = None;
+3
View File
@@ -35418,6 +35418,9 @@ components:
type: string
last_pull_status:
$ref: "#/components/schemas/AutoPullStatus"
enabled_by:
type: string
description: Email of the admin automatic pulls apply changes as. Set by the server when the settings are saved.
required:
- enabled
@@ -1587,10 +1587,10 @@ pub(crate) async fn tarball_workspace(
// Use v2 format only if explicitly requested, otherwise use v1 (legacy) for backward compatibility
// Server-owned state (the HMAC webhook secret + hook id/error, the
// synced-sha / last-pull status, and what the credential check observed)
// must never leave the server: keep it out of export archives and synced
// repos, and don't let a re-imported workspace inherit another install's
// hook/sync state. Mirrors the GET-settings redaction.
// synced-sha / last-pull status, the admin automatic pulls run as, and what
// the credential check observed) must never leave the server: keep it out of
// export archives and synced repos, and don't let a re-imported workspace
// inherit another install's hook/sync state or pull identity.
fn redact_git_sync_for_export(git_sync: Option<Value>) -> Option<Value> {
let mut git_sync = git_sync?;
if let Some(repos) = git_sync
@@ -1607,6 +1607,7 @@ pub(crate) async fn tarball_workspace(
"webhook_error",
"last_synced_sha",
"last_pull_status",
"enabled_by",
] {
auto_pull.remove(field);
}
@@ -514,6 +514,11 @@ pub struct AutoPullSettings {
pub last_synced_sha: std::collections::HashMap<String, String>,
#[serde(default, skip_serializing_if = "Option::is_none")]
pub last_pull_status: Option<AutoPullStatus>,
/// Email of the admin this repository's automatic pulls (its own and its forks')
/// apply changes as: whoever last saved the settings with auto pull on. Stamped
/// server-side, never taken from the client.
#[serde(default, skip_serializing_if = "Option::is_none")]
pub enabled_by: Option<String>,
}
// Manual Debug so the HMAC `webhook_secret` (even encrypted) never lands in logs.
@@ -524,6 +529,7 @@ impl std::fmt::Debug for AutoPullSettings {
.field("mode", &self.mode)
.field("poll_interval_s", &self.poll_interval_s)
.field("sync_forks", &self.sync_forks)
.field("enabled_by", &self.enabled_by)
.field("webhook_id", &self.webhook_id)
.field(
"webhook_secret",
@@ -2709,6 +2715,7 @@ mod tests {
webhook_secret: None,
webhook_url: None,
webhook_error: None,
enabled_by: None,
last_synced_sha: synced
.iter()
.map(|(r, s)| (r.to_string(), s.to_string()))
+3 -2
View File
@@ -9,8 +9,8 @@ name = "windmill_git_sync"
path = "./src/lib.rs"
[features]
private = ["windmill-common/private", "windmill-dep-map/private"]
enterprise = ["windmill-queue/enterprise", "windmill-common/enterprise"]
private = ["windmill-common/private", "windmill-dep-map/private", "windmill-audit/private"]
enterprise = ["windmill-queue/enterprise", "windmill-common/enterprise", "windmill-audit/enterprise"]
all_sqlx_features = ["enterprise"]
default = []
@@ -23,5 +23,6 @@ tracing.workspace = true
windmill-common = { workspace = true, default-features = false }
windmill-queue.workspace = true
windmill-dep-map.workspace = true
windmill-audit.workspace = true
regex = "1.10.3"
tokio = { workspace = true, features = ["full"] }