Remove em-dashes: the rest

Author: vdavid

.claude/commands/execute.md

@@ -18,7 +18,7 @@ Agents:
   whole plan. This usually results in a disaster in quality because they run out of context, they auto-compress, then
   the compressed agent lacks proper understanding of our values and what we're doing. So again, give them a clear scope.
 - Ask every agent to reflect whether they are satisfied with what they'd done. Make them ask: "Is what I've done solid
-  AND elegant? You proud and confident about it?" — If the answer is "no" to either, they should adjust, then rinse and
+  AND elegant? You proud and confident about it?" If the answer is "no" to either, they should adjust, then rinse and
   repeat.
 - They should also see if there is something else to fix, like any latent bugs that only need 10–15 LoC changes around
   their part of the development. They're encouraged to fix these. Correctness and bug-free code over crystal-clean

.claude/commands/release.md

@@ -3,15 +3,15 @@ Prepare a release based on docs/guides/releasing.md.
 1. Prerequisite: Run `gh secret list` and verify that `TAURI_SIGNING_PRIVATE_KEY` and
    `TAURI_SIGNING_PRIVATE_KEY_PASSWORD` both exist. If either is missing, warn the user and stop.
 2. Update @CHANGELOG.md based on git commits since last release.
-   - Commits have title + body — read all!
+   - Commits have title + body. Read all!
    - You can link multiple commits for changelog items if needed.
    - List major but non-app changes in a "Non-app" section.
-   - **Get commit SHAs via `git log --format='%h' --abbrev=8`** — never extend a 7-char prefix from `git log --oneline`
+   - **Get commit SHAs via `git log --format='%h' --abbrev=8`**. Never extend a 7-char prefix from `git log --oneline`
      by guessing the next character. The committed changelog convention is 8 chars; let git produce them. The
      `changelog-links` check will reject fabricated SHAs and abort the release.
    - **Add a `## [Unreleased]` heading** right after the format preamble (before the first versioned section), then put
      entries under it. The release script replaces this heading with the versioned one. The committed changelog has no
-     `[Unreleased]` section between releases — you're creating it fresh each time.
+     `[Unreleased]` section between releases. You're creating it fresh each time.
 
    ### Style: plain-sentence, dense, impact-focused
    - **Each entry is one sentence.** No `**Bold title:** Body.`; the headline IS the entry. Most entries land under 20
@@ -59,7 +59,7 @@ Prepare a release based on docs/guides/releasing.md.
 
    > - Add friendly errors for git browser ([19d5b075](...), [af64689f](...)).
 
-   **Keep long when warranted** (true tentpole launches like Linux alpha — many real commits, big story):
+   **Keep long when warranted** (true tentpole launches like Linux alpha, with many real commits and a big story):
 
    > - Add Linux support (alpha): volumes via /proc/mounts, file ops with reflink support, trash via FreeDesktop spec,
    >   inotify file watching, MTP ungated, SMB via mDNS + smbclient fallback, GVFS-mounted shares as volumes, native
@@ -76,18 +76,18 @@ Prepare a release based on docs/guides/releasing.md.
 7. **After pushing**, confirm the self-hosted runner picked up the build:
    - Wait ~30 seconds, then run `gh run view <release-run-id> --json jobs` and check the `Build (...)` jobs.
    - At least one `Build (...)` job should be `in_progress` (the self-hosted runner serializes the three matrix jobs, so
-     the others stay `queued` — that's normal).
+     the others stay `queued`, which is normal).
    - **If all three are still `queued` after ~30s, the self-hosted runner is down.** Confirm with
      `launchctl list | grep cmdr` and look for `actions.runner.vdavid-cmdr.*`. Restart with
      `cd ~/actions-runner-cmdr && ./svc.sh start` (fall back to `launchctl bootout` + `bootstrap` if `svc.sh` errors
      with "Load failed: 5: Input/output error"). Re-check after another 30 s. The queued jobs pick up automatically once
-     the runner reports in — no need to re-trigger or re-tag.
+     the runner reports in. No need to re-trigger or re-tag.
 8. **Then arm `caffeinate`** to prevent the Mac from sleeping during the build. The self-hosted runner lives on this
-   Mac; any sleep — display or system — drops the runner connection and fails every in-flight matrix job with
+   Mac; any sleep (display or system) drops the runner connection and fails every in-flight matrix job with
    `The self-hosted runner lost communication with the server`. See `docs/guides/releasing.md` § "Keep the Mac awake
    during the build".
    - Run `caffeinate -dimsu` as a Bash `run_in_background` call. Capture the background task id so you can stop it.
-   - Disarm it once the release workflow reports `completed` (success or failure — not just when the matrix is done).
+   - Disarm it once the release workflow reports `completed` (success or failure, not just when the matrix is done).
    - If the user requests a re-run of failed jobs, re-arm caffeinate first.
 9. **Monitor the CI build**:
    - Remind the user not to close their laptop for ~15 minutes while the self-hosted runner builds.
@@ -97,14 +97,14 @@ Prepare a release based on docs/guides/releasing.md.
      fix.
    - Suggest the user to also track the build at https://github.com/vdavid/cmdr/actions.
 10. **In parallel, watch the standalone CI run** (the non-release `CI` workflow that fires on the same push):
-    - It's not a blocker for the release. If it goes red, fix it in the background while the release builds — small
+    - It's not a blocker for the release. If it goes red, fix it in the background while the release builds. Small
       things like lint regressions are common.
     - Surface the failure to the user when convenient; don't interrupt release-build progress reporting for it.
 11. **After the release run succeeds, verify the public surface**:
-    - `gh release view vX.Y.Z --json assets,tagName,publishedAt` — confirm the expected DMGs are attached
+    - `gh release view vX.Y.Z --json assets,tagName,publishedAt`: confirm the expected DMGs are attached
       (`Cmdr_X.Y.Z_aarch64.dmg`, `_x64.dmg`, `_universal.dmg`) and sizes look reasonable.
     - Wait ~30 seconds for the website auto-deploy (the release workflow commits an updated `latest.json` and fires a
       webhook), then `curl -s https://getcmdr.com/latest.json | jq -r .version` and confirm it matches `X.Y.Z`.
     - If `latest.json` still shows the old version after ~2 minutes, the deploy webhook may have failed silently. Tell
       the user; the manual fix is to re-trigger the website-deploy workflow via `workflow_dispatch` from the Actions
-      tab. Don't block release success on this — the GitHub Release is what users actually download.
+      tab. Don't block release success on this. The GitHub Release is what users actually download.

.github/workflows/release.yml

@@ -77,7 +77,7 @@ jobs:
 
       - name: Detach stale Cmdr* DMG mounts
         # tauri-action's bundle_dmg.sh fails fast (~3 s) when the "Cmdr" volume name is already
-        # taken in /Volumes (typically from a previous Finder double-click on an old DMG —
+        # taken in /Volumes (typically from a previous Finder double-click on an old DMG;
         # the self-hosted runner shares /Volumes with the dev's interactive session). Detach any
         # leftovers so the new bundle gets the canonical volume name. scripts/release.sh has the
         # same guard pre-tag-push, but this catches mounts that appear AFTER the tag push.

apps/desktop/scripts/e2e-linux.sh

@@ -262,7 +262,7 @@ SMB_E2E_SERVICES=(smb-consumer-guest smb-consumer-auth smb-consumer-50shares smb
 
 # probe_smb_ports returns 0 if every required service's published port 445
 # accepts TCP within $1 seconds, otherwise 1. NEVER replace this with a
-# blanket `sleep N` — see apps/desktop/test/CLAUDE.md "Testing principles".
+# blanket `sleep N`; see apps/desktop/test/CLAUDE.md "Testing principles".
 probe_smb_ports() {
     local timeout="${1:-10}"
     local deadline=$((SECONDS + timeout))
@@ -292,7 +292,7 @@ start_smb_containers() {
     # the other two fail with "Cannot reach smb-consumer-50shares".
     #
     # ALSO: "running" per `docker compose ps` only means the container is
-    # alive — smbd inside may be hung, OOM-killed, or still loading. We always
+    # alive; smbd inside may be hung, OOM-killed, or still loading. We always
     # follow the running-check with an active TCP probe; if it fails, we
     # restart the SMB stack rather than letting the E2E run hit "Cannot reach"
     # errors mid-test. See the case study in
@@ -308,11 +308,11 @@ start_smb_containers() {
     done
 
     if $all_running; then
-        log_info "SMB containers already running — verifying smbd reachability..."
+        log_info "SMB containers already running; verifying smbd reachability..."
         if probe_smb_ports 10; then
             log_info "SMB containers healthy"
         else
-            log_warn "SMB containers running but not serving — restarting"
+            log_warn "SMB containers running but not serving; restarting"
             docker compose -p smb-consumer down > /dev/null 2>&1 || true
             "$SMB_SERVERS_DIR/start.sh" e2e
         fi
@@ -337,7 +337,7 @@ start_smb_containers() {
     if probe_smb_ports 30; then
         log_info "SMB e2e stack ready: all 4 containers accepting TCP on :445"
     else
-        log_error "SMB e2e stack NOT ready after restart — aborting before tests"
+        log_error "SMB e2e stack NOT ready after restart; aborting before tests"
         docker compose -p smb-consumer ps
         for service in "${SMB_E2E_SERVICES[@]}"; do
             log_warn "--- last 30 lines of $service log ---"
@@ -475,7 +475,7 @@ else
     if probe_smb_ports 5; then
         log_info "SMB post-flight: all 4 containers still accepting TCP on :445"
     else
-        log_warn "SMB post-flight: at least one container is no longer accepting TCP — likely died mid-run"
+        log_warn "SMB post-flight: at least one container is no longer accepting TCP, likely died mid-run"
         for service in "${SMB_E2E_SERVICES[@]}"; do
             state=$(docker compose -p smb-consumer ps --format '{{.State}} {{.Status}}' "$service" 2>/dev/null | head -1)
             log_warn "  $service: ${state:-unknown}"

apps/desktop/src-tauri/src/commands/rename.rs

@@ -253,7 +253,7 @@ fn check_macos_flags(path: &Path) -> Result<(), String> {
     // SAFETY: c_path is valid, stat is a valid pointer
     let result = unsafe { libc::lstat(c_path.as_ptr(), stat.as_mut_ptr()) };
     if result != 0 {
-        // Can't stat — file may have been deleted, let the rename itself fail with a clear error
+        // Can't stat; file may have been deleted, let the rename itself fail with a clear error
         return Ok(());
     }
 
@@ -409,7 +409,7 @@ async fn check_sibling_conflict_via_volume(volume_id: &str, new_path: &Path) ->
 
     let entry = match volume.get_metadata(new_path).await {
         Ok(e) => e,
-        Err(_) => return (false, None), // No conflict — file doesn't exist
+        Err(_) => return (false, None), // No conflict: file doesn't exist
     };
 
     let conflict = ConflictFileInfo {
@@ -456,7 +456,7 @@ mod tests {
     async fn test_check_rename_permission_nonexistent() {
         let result = check_rename_permission("/nonexistent_12345/file.txt".to_string()).await;
         assert!(result.is_err());
-        // allowed-error-string-match: IpcError is a flat struct with no typed variants — message content is the only signal
+        // allowed-error-string-match: IpcError is a flat struct with no typed variants; message content is the only signal
         assert!(result.unwrap_err().message.contains("doesn't exist"));
     }
 
@@ -580,7 +580,7 @@ mod tests {
         )
         .await;
         assert!(result.is_err());
-        // allowed-error-string-match: IpcError is a flat struct with no typed variants — message content is the only signal
+        // allowed-error-string-match: IpcError is a flat struct with no typed variants; message content is the only signal
         assert!(result.unwrap_err().message.contains("already exists"));
         // Both files still intact
         assert!(old.exists());
@@ -629,7 +629,7 @@ mod tests {
     async fn test_move_to_trash_nonexistent() {
         let result = move_to_trash("/nonexistent_12345/trash_me.txt".to_string()).await;
         assert!(result.is_err());
-        // allowed-error-string-match: IpcError is a flat struct with no typed variants — message content is the only signal
+        // allowed-error-string-match: IpcError is a flat struct with no typed variants; message content is the only signal
         assert!(result.unwrap_err().message.contains("doesn't exist"));
     }
 }

apps/desktop/src-tauri/src/file_system/write_operations/move_op.rs

@@ -20,7 +20,7 @@ use super::types::{
 // ============================================================================
 
 /// Tracks renames performed during same-FS move for rollback on cancellation.
-/// Each entry is `(original_source, moved_to_dest)` — rollback reverses them.
+/// Each entry is `(original_source, moved_to_dest)`. Rollback reverses them.
 struct MoveTransaction {
     renames: Vec<(PathBuf, PathBuf)>,
 }
@@ -154,7 +154,7 @@ fn move_with_rename(
                     }
                 }
             } else {
-                // No conflict — just rename
+                // No conflict, so just rename
                 fs::rename(source, &dest_path).with_path(source)?;
                 move_tx.record(source.clone(), dest_path);
             }
@@ -223,7 +223,7 @@ fn move_with_rename(
 /// The two look similar in structure but differ in every detail (copy has progress tracking,
 /// symlink handling, byte counting, transaction recording, strategy selection). A shared
 /// abstraction would be forced and fragile. See `copy.rs` `copy_single_item` for the copy side.
-#[allow(clippy::too_many_arguments, reason = "intentional — see doc comment above")]
+#[allow(clippy::too_many_arguments, reason = "intentional; see doc comment above")]
 fn merge_move_directory(
     source_dir: &Path,
     dest_dir: &Path,
@@ -253,7 +253,7 @@ fn merge_move_directory(
         }
 
         if source_child.is_dir() && dest_child.exists() && dest_child.is_dir() {
-            // Both are directories — recurse
+            // Both are directories, recurse
             merge_move_directory(
                 &source_child,
                 &dest_child,
@@ -280,12 +280,12 @@ fn merge_move_directory(
                     move_tx.record(source_child, resolved.path);
                 }
                 None => {
-                    // Skip — source file stays in place
+                    // Skip: source file stays in place
                     continue;
                 }
             }
         } else {
-            // No conflict — just rename
+            // No conflict, just rename
             fs::rename(&source_child, &dest_child).with_path(&source_child)?;
             move_tx.record(source_child, dest_child);
         }
@@ -435,7 +435,7 @@ fn move_with_staging(
             let final_path = destination.join(file_name);
 
             // When both staged and final are directories, merge recursively.
-            // No MoveTransaction needed here — staging cleanup handles rollback.
+            // No MoveTransaction needed here: staging cleanup handles rollback.
             let mut staging_move_tx = MoveTransaction::new();
             if staged_path.is_dir() && final_path.exists() && final_path.is_dir() {
                 merge_move_directory(
@@ -476,7 +476,7 @@ fn move_with_staging(
                     }
                 }
             } else {
-                // No conflict — just rename from staging to final
+                // No conflict, just rename from staging to final
                 fs::rename(&staged_path, &final_path).map_err(|e| WriteOperationError::IoError {
                     path: staged_path.display().to_string(),
                     message: format!("Failed to move from staging: {}", e),

apps/desktop/src-tauri/src/indexing/aggregator.rs

@@ -838,7 +838,7 @@ mod tests {
             ],
         );
 
-        // Only compute stats for /a/sub (id=4) — leave /a (id=2) and root (id=1) missing
+        // Only compute stats for /a/sub (id=4): leave /a (id=2) and root (id=1) missing
         IndexStore::upsert_dir_stats_by_id(
             &conn,
             &[DirStatsById {
@@ -894,7 +894,7 @@ mod tests {
         //   /grand/parent (id=3)
         //   /grand/parent/leaf (id=4)
         //   /grand/parent/leaf/link (id=5, symlink)
-        //   /grand/sibling (id=6) — no symlinks
+        //   /grand/sibling (id=6), no symlinks
         //   /grand/sibling/file.txt (id=7, 100 bytes)
         insert_entries(
             &conn,
@@ -955,7 +955,7 @@ mod tests {
     #[test]
     fn aggregate_dir_with_only_symlinks_has_zero_size() {
         let (conn, _dir) = open_temp_conn();
-        // /links contains only two symlinks — total size 0, but flag is true
+        // /links contains only two symlinks: total size 0, but flag is true
         insert_entries(
             &conn,
             &[
@@ -1053,7 +1053,7 @@ mod tests {
                             id, cp, pid, pp
                         );
                     }
-                    // Transitively the same must hold for any ancestor — but
+                    // Transitively the same must hold for any ancestor,
                     // chain through `parent_of` to be sure.
                     let mut cursor = pid;
                     let mut hops = 0;