Skip to content

Worktrunk is a CLI for Git worktree management, designed for parallel AI agent workflows

fix(config): migrate `[ci] platform` into a `[forge]` table that lacks it (#4061)

Worktrunk Bot3d1b5d58b1b3
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 020568b3c..6131391d6 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -12,6 +12,8 @@
### Fixed
+- **`wt config update` migrates `[ci] platform` into a `[forge]` section that only sets `hostname`**: the migration stood down whenever a `[forge]` section existed at all, so the config a GitHub Enterprise or self-hosted GitLab user ends up with — `[ci] platform` from before the rename, `[forge] hostname` added later for an SSH host alias — kept the deprecated key with no warning and nothing `wt config update` would do about it. The platform still resolved, so nothing broke; the deprecation notice that precedes `[ci]`'s eventual removal simply never arrived. A `[forge]` that already sets `platform`, or a `forge` that isn't a section, still stands down.
+
- **`wt config shell install` migrates a fish wrapper at the deprecated `conf.d` path even when `~/.config/fish/functions` doesn't exist yet**: fish was skipped for want of a config location, so the bare command left the deprecated wrapper running and only `wt config shell install fish` migrated it. A worktrunk wrapper at the old path now counts as fish being configured, just at the old path. `wt config show` reports that wrapper too, where before it showed nothing at all for fish unless fish was on `PATH`.
- **`wt config shell install` no longer tells an already-wrapped shell to restart**: reinstalling from inside a shell that has the wrapper loaded — after a version bump, or when the fish extension moves from `conf.d` to `functions` — printed `↳ Restart shell to activate shell integration` right after the wrapper had intercepted the command. The hint now only appears when integration isn't active.
diff --git a/src/config/deprecation.rs b/src/config/deprecation.rs
index aec9a5135..72b809821 100644
--- a/src/config/deprecation.rs
+++ b/src/config/deprecation.rs
@@ -555,8 +555,9 @@ enum RulePass {
/// earlier rules have already migrated, so a rule reports a pattern in its
/// post-migration shape.
///
-/// The `[ci]` → `[forge]` rule is order-independent: `[forge]` takes over
-/// `[ci]`'s explicit document position (see [`migrate_ci_doc`]), so its
+/// The `[ci]` → `[forge]` rule is order-independent: `[forge]` renders at an
+/// explicit document position either way — the one it was parsed at, or
+/// `[ci]`'s when the rule creates the table (see [`migrate_ci_doc`]) — so its
/// rendered placement doesn't depend on which tables other rules re-append.
///
/// A [`DeprecationRule::Structural`] rule must not depend on an `UpdateOnly`
@@ -1081,12 +1082,18 @@ fn table_like_len(item: &toml_edit::Item) -> Option<usize> {
/// rewrite could meaningfully move. Removes `[ci]` if `platform` was its only
/// field.
///
-/// An existing `forge` key of any shape suppresses the migration: a `[forge]`
-/// table means the user already migrated, and overwriting a malformed scalar
-/// `forge = "x"` would silently drop their config — serde's type error points
-/// at it instead.
+/// What suppresses the migration is an occupied *destination key*, not the
+/// presence of `forge`: a `[forge]` that already carries `platform` means the
+/// user already migrated, and a `forge` that isn't a table (`forge = "x"`) has
+/// no key to insert into — overwriting it would silently drop their config,
+/// where serde's type error points at it instead. A `[forge]` table without
+/// `platform` is the common half-migrated shape (`hostname` set for a GHE or
+/// self-hosted GitLab remote, `platform` still back in `[ci]`), and its slot is
+/// free, so `platform` lands there. Suppressing on the table's mere existence
+/// left that user un-migrated *and* unwarned, so the deprecated key kept
+/// working silently — right up until `[ci]` is removed.
///
-/// `[forge]` takes over `[ci]`'s document position — a fresh table has no
+/// A fresh `[forge]` takes over `[ci]`'s document position — a new table has no
/// position and would render at the end of the file instead of in the user's
/// original spot. The `platform` entry moves wholesale (key and item), so
/// comments attached to the line survive. When `[ci]` is fully consumed, its
@@ -1094,10 +1101,26 @@ fn table_like_len(item: &toml_edit::Item) -> Option<usize> {
/// when other keys keep `[ci]` alive, the decor stays there and `[forge]`
/// renders directly after the remainder — it shares `[ci]`'s position, the
/// position sort is stable, and `[forge]` is inserted later in visit order.
+///
+/// Inserting into an *existing* `[forge]` has no position to take over and no
+/// home for `[ci]`'s decor: `[forge]` is wherever the user wrote it, which can
+/// be far from the comment above `[ci]`. So an emptied `[ci]` is removed only
+/// when its own decor is blank — a commented one stays as an empty section,
+/// which contributes no config, raises no warning, and keeps the user's prose
+/// where they put it.
fn migrate_ci_doc(doc: &mut toml_edit::DocumentMut) -> Deprecations {
- if doc.get("forge").is_some() {
- return Vec::new();
- }
+ // Read the destination before touching `[ci]`, so a suppressed migration
+ // leaves the document byte-identical.
+ let forge_slot = match doc.get("forge") {
+ None => ForgeSlot::Absent,
+ Some(forge) => match forge.as_table() {
+ Some(table) if !table.contains_key("platform") => ForgeSlot::FreeSlot,
+ // `platform` already there, or `forge` is a scalar, an array, or an
+ // inline table we can't extend without reformatting what the user
+ // wrote.
+ _ => return Vec::new(),
+ },
+ };
let Some(ci_table) = doc.get_mut("ci").and_then(|ci| ci.as_table_mut()) else {
return Vec::new();
@@ -1114,19 +1137,57 @@ fn migrate_ci_doc(doc: &mut toml_edit::DocumentMut) -> Deprecations {
let (key, item) = ci_table
.remove_entry("platform")
.expect("checked platform exists above");
- let mut forge_table = toml_edit::Table::new();
- forge_table.insert_formatted(&key, item);
- forge_table.set_position(ci_table.position());
- if ci_table.is_empty() {
- *forge_table.decor_mut() = ci_table.decor().clone();
- doc.remove("ci");
+ let ci_emptied = ci_table.is_empty();
+ let ci_position = ci_table.position();
+ let ci_decor = ci_table.decor().clone();
+
+ match forge_slot {
+ ForgeSlot::Absent => {
+ let mut forge_table = toml_edit::Table::new();
+ forge_table.insert_formatted(&key, item);
+ forge_table.set_position(ci_position);
+ if ci_emptied {
+ *forge_table.decor_mut() = ci_decor;
+ doc.remove("ci");
+ }
+ doc.insert("forge", toml_edit::Item::Table(forge_table));
+ }
+ ForgeSlot::FreeSlot => {
+ let forge_table = doc
+ .get_mut("forge")
+ .and_then(|forge| forge.as_table_mut())
+ .expect("checked forge is a table above");
+ forge_table.insert_formatted(&key, item);
+ // A `[forge]` that exists only because of a sub-table renders no
+ // header of its own until something is written directly into it.
+ forge_table.set_implicit(false);
+ if ci_emptied && decor_prefix_is_blank(&ci_decor) {
+ doc.remove("ci");
+ }
+ }
}
- doc.insert("forge", toml_edit::Item::Table(forge_table));
-
vec![DeprecationKind::CiSection]
}
+/// Where the migrated `platform` key is headed — see [`migrate_ci_doc`].
+enum ForgeSlot {
+ /// No `forge` key at all: build the table and take over `[ci]`'s position.
+ Absent,
+ /// A `[forge]` table whose `platform` slot is free: insert in place.
+ FreeSlot,
+}
+
+/// Whether a table's leading decor carries nothing worth keeping (no comment,
+/// just whitespace). Decor that was never parsed from a document reads as
+/// blank — it holds no user text either way.
+fn decor_prefix_is_blank(decor: &toml_edit::Decor) -> bool {
+ decor
+ .prefix()
+ .and_then(|prefix| prefix.as_str())
+ .is_none_or(|prefix| prefix.trim().is_empty())
+}
+
/// Migrate a negated boolean field within a table (e.g., `no-ff = true` →
/// `ff = false`).
///
@@ -3269,7 +3330,7 @@ ff = false
"#);
}
- /// A `forge` key of any shape suppresses the `[ci]` migration — a scalar
+ /// A `forge` that isn't a table suppresses the `[ci]` migration — a scalar
/// `forge = "x"` must not be overwritten by the migrated table (serde's
/// type error points at it instead).
#[test]
@@ -3286,6 +3347,75 @@ json-schema = 1
assert!(detect_deprecations(content).is_empty());
}
+ /// A `[forge]` table that only sets `hostname` has the `platform` slot
+ /// free, so the deprecated key moves into it rather than being left behind.
+ /// This is the shape a self-hosted GitLab or GHE user lands in by adding
+ /// `[forge] hostname` to a config that already carried `[ci] platform`;
+ /// suppressing on the table's mere existence left them un-migrated and
+ /// unwarned. `[ci]` was consumed entirely, so it goes.
+ #[test]
+ fn test_ci_migration_fills_free_slot_in_existing_forge() {
+ let content = r#"[ci]
+platform = "gitlab"
+
+[forge]
+hostname = "gitlab.example.com"
+"#;
+ assert!(matches!(
+ detect_deprecations(content).as_slice(),
+ [DeprecationKind::CiSection]
+ ));
+ insta::assert_snapshot!(migrate_content(content), @r#"
+ [forge]
+ hostname = "gitlab.example.com"
+ platform = "gitlab"
+ "#);
+ }
+
+ /// An existing `[forge]` is wherever the user wrote it, so there is nowhere
+ /// to carry a comment sitting above `[ci]`. An emptied but commented `[ci]`
+ /// stays as an empty section — it contributes no config and raises no
+ /// warning, and dropping it would drop the user's prose with it.
+ #[test]
+ fn test_ci_migration_keeps_commented_ci_when_forge_exists() {
+ let content = r#"# talk to the internal instance
+[ci]
+platform = "gitlab"
+
+[forge]
+hostname = "gitlab.example.com"
+"#;
+ insta::assert_snapshot!(migrate_content(content), @r#"
+ # talk to the internal instance
+ [ci]
+
+ [forge]
+ hostname = "gitlab.example.com"
+ platform = "gitlab"
+ "#);
+ }
+
+ /// A surviving `[ci]` key keeps the section either way; only `platform`
+ /// moves, and it lands in the existing `[forge]` rather than a new one.
+ #[test]
+ fn test_ci_migration_into_existing_forge_preserves_other_keys() {
+ let content = r#"[ci]
+platform = "gitea"
+hostname = "ci.example"
+
+[forge]
+hostname = "forge.example"
+"#;
+ insta::assert_snapshot!(migrate_content(content), @r#"
+ [ci]
+ hostname = "ci.example"
+
+ [forge]
+ hostname = "forge.example"
+ platform = "gitea"
+ "#);
+ }
+
/// The framework invariant: a warning fires exactly when `wt config
/// update` would change the file. Degenerate configs that can't be safely
/// rewritten produce no warning and no rewrite; deprecated configs
@@ -3300,9 +3430,10 @@ json-schema = 1
"[ci]\nplatform = \"\"\n",
"[ci]\nplatform = 42\n",
"[ci]\nhostname = \"ghe.example\"\n",
- // forge already exists — table or malformed scalar
+ // the destination key is occupied, or `forge` is not a table
"[forge]\nplatform = \"gitlab\"\n\n[ci]\nplatform = \"github\"\n",
"forge = \"x\"\n\n[ci]\nplatform = \"github\"\n",
+ "forge = { hostname = \"ghe.example\" }\n\n[ci]\nplatform = \"github\"\n",
// empty deprecated sections contribute no config
"[commit-generation]\n",
"[select]\n",
@@ -3336,6 +3467,10 @@ json-schema = 1
let rewritten = [
"[ci]\nplatform = \"github\"\n",
+ // a `[forge]` table whose `platform` slot is free is a destination,
+ // not a blocker — with and without a comment holding `[ci]` open
+ "[ci]\nplatform = \"github\"\n\n[forge]\nhostname = \"ghe.example\"\n",
+ "# internal\n[ci]\nplatform = \"github\"\n\n[forge]\nhostname = \"ghe.example\"\n",
"[merge]\nff = true\nno-ff = true\n",
// negated bool written inline (`merge = { no-ff = true }`) migrates
// like the section form