ci: build via host docker socket (plain build); fix missing daemon socket

build-and-push failed at docker/setup-buildx-action: the job had no
/var/run/docker.sock, so buildx's docker-container driver couldn't reach
the daemon. Mount the host socket into build-and-push and deploy, and
replace setup-buildx + build-push-action (+ the unsupported gha cache)
with a plain docker build/push against the host daemon (DooD), reusing
the host's layer cache.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

.gitea/workflows/deploy.yml

@@ -72,9 +72,17 @@ jobs:
     runs-on: ubuntu-latest
     needs: [test-backend, test-frontend]
     # PRs only run the test jobs; build + deploy are reserved for
-    # post-merge pushes to main. Without this gate every PR would push
+    # post-merge pushes to main.
-    # a tagged image to the registry and SSH-deploy to prod.
     if: github.event_name != 'pull_request'
+    # Build on the host docker daemon directly (docker-outside-of-docker):
+    # the runner shares the deploy host's daemon, so a plain `docker build`
+    # reuses the host's layer cache and avoids buildx's docker-container
+    # driver + the gha cache exporter — neither works against this single-host
+    # act_runner, and there is no in-job daemon socket unless we mount it.
+    container:
+      image: docker.gitea.com/runner-images:ubuntu-latest
+      volumes:
+        - /var/run/docker.sock:/var/run/docker.sock
     outputs:
       image_tag: ${{ steps.meta.outputs.image_tag }}
       version: ${{ steps.meta.outputs.version }}
@@ -93,48 +101,32 @@ jobs:
           echo "image_tag=${GITHUB_SHA}" >> "$GITHUB_OUTPUT"
           echo "version=${version}" >> "$GITHUB_OUTPUT"
 
-      - uses: docker/setup-buildx-action@v3
+      - name: Build & push backend + frontend
-
+        env:
-      - name: docker login
+          REGISTRY_URL: ${{ secrets.REGISTRY_URL }}
-        uses: docker/login-action@v3
+          REGISTRY_USERNAME: ${{ secrets.REGISTRY_USERNAME }}
-        with:
+          REGISTRY_PASSWORD: ${{ secrets.REGISTRY_PASSWORD }}
-          registry: ${{ secrets.REGISTRY_URL }}
+          IMAGE_TAG: ${{ steps.meta.outputs.image_tag }}
-          username: ${{ secrets.REGISTRY_USERNAME }}
+          VERSION: ${{ steps.meta.outputs.version }}
-          password: ${{ secrets.REGISTRY_PASSWORD }}
+        run: |
-
+          set -eu
-      - name: Build & push backend
+          echo "$REGISTRY_PASSWORD" | docker login "$REGISTRY_URL" -u "$REGISTRY_USERNAME" --password-stdin
-        uses: docker/build-push-action@v5
+          for svc in backend frontend; do
-        with:
+            img="$REGISTRY_URL/mangalord-$svc"
-          context: ./backend
+            docker build -t "$img:$IMAGE_TAG" -t "$img:latest" -t "$img:$VERSION" "./$svc"
-          push: true
+            for tag in "$IMAGE_TAG" latest "$VERSION"; do docker push "$img:$tag"; done
-          tags: |
+          done
-            ${{ secrets.REGISTRY_URL }}/mangalord-backend:latest
+          docker logout "$REGISTRY_URL"
-            ${{ secrets.REGISTRY_URL }}/mangalord-backend:${{ steps.meta.outputs.image_tag }}
-            ${{ secrets.REGISTRY_URL }}/mangalord-backend:${{ steps.meta.outputs.version }}
-          cache-from: type=gha,scope=backend
-          cache-to: type=gha,mode=max,scope=backend
-
-      - name: Build & push frontend
-        uses: docker/build-push-action@v5
-        with:
-          context: ./frontend
-          push: true
-          tags: |
-            ${{ secrets.REGISTRY_URL }}/mangalord-frontend:latest
-            ${{ secrets.REGISTRY_URL }}/mangalord-frontend:${{ steps.meta.outputs.image_tag }}
-            ${{ secrets.REGISTRY_URL }}/mangalord-frontend:${{ steps.meta.outputs.version }}
-          cache-from: type=gha,scope=frontend
-          cache-to: type=gha,mode=max,scope=frontend
 
   deploy:
     runs-on: ubuntu-latest
     needs: build-and-push
     if: github.event_name != 'pull_request'
     # Single-host deploy: the runner lives on the same box as the stack, so we
-    # drive the host docker daemon directly (act_runner shares its socket via
+    # drive the host docker daemon directly (the job mounts the host docker
-    # `docker_host: "-"`) instead of SSHing out. The compose dir is bind-mounted
+    # socket) instead of SSHing out. The compose dir is bind-mounted at its
-    # at its REAL host path so compose's relative bind-mounts (./mangalord/...,
+    # REAL host path so compose's relative bind-mounts (./mangalord/...,
-    # ./Caddyfile) resolve; this requires `/mnt/ssd/docker-data` in the runner's
+    # ./Caddyfile) resolve; both paths must be in the runner's
     # container.valid_volumes. The central compose references the images as
     # registry.mc02.dev/mangalord-*:${MANGALORD_TAG:-latest}, so we only pull
     # and recreate the two mangalord services at the freshly built SHA.
@@ -142,6 +134,7 @@ jobs:
       image: docker:cli
       volumes:
         - /mnt/ssd/docker-data:/mnt/ssd/docker-data
+        - /var/run/docker.sock:/var/run/docker.sock
     steps:
       - name: Deploy to the local stack
         working-directory: /mnt/ssd/docker-data

backend/Cargo.lock

@@ -1470,7 +1470,7 @@ checksum = "c41e0c4fef86961ac6d6f8a82609f55f31b05e4fce149ac5710e439df7619ba4"
 
 [[package]]
 name = "mangalord"
-version = "0.45.0"
+version = "0.45.1"
 dependencies = [
  "anyhow",
  "argon2",

backend/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "mangalord"
-version = "0.45.0"
+version = "0.45.1"
 edition = "2021"
 default-run = "mangalord"
 

backend/src/crawler/source/target.rs

@@ -7,7 +7,6 @@
 //! (`td:has(label:contains("Author:"))`) are implemented by walking
 //! the parsed tree.
 
-use std::collections::VecDeque;
 use std::time::Duration;
 
 use anyhow::Context;
@@ -75,10 +74,11 @@ impl Source for TargetSource {
         &self,
         ctx: &FetchContext<'_>,
     ) -> anyhow::Result<Box<dyn DiscoverWalk + Send>> {
-        // Always visit page 1 first because that's the only way to
+        // Probe page 1 up front (with transient retry) for two reasons:
-        // discover `last_page`. Retry it on transient — a broken first
+        // a broken first page should abort cleanly rather than mid-walk,
-        // page would otherwise abort the whole walk before we've even
+        // and the HTML is handed straight to the first `next_batch` call
-        // started.
+        // so the walker doesn't re-fetch it. Page count is discovered
+        // incrementally — see `TargetSourceWalker::next_batch`.
         let first_html = retry_on_transient(
             || async {
                 navigate(ctx, self.base_url.as_str(), LIST_PAGE_MARKER).await
@@ -87,21 +87,10 @@ impl Source for TargetSource {
             PAGE_TRANSIENT_RETRY_DELAY,
         )
         .await?;
-        let last_page = {
-            let doc = scraper::Html::parse_document(&first_html);
-            parse_last_page(&doc)
-        };
-
-        let order = build_page_order(last_page);
-        tracing::info!(
-            last_page = ?last_page,
-            page_count = order.len(),
-            "walking pagination"
-        );
 
         Ok(Box::new(TargetSourceWalker {
             base_url: self.base_url.clone(),
-            pages_remaining: order,
+            next_page: 1,
             first_page_html: Some(first_html),
         }))
     }
@@ -147,24 +136,19 @@ impl Source for TargetSource {
     }
 }
 
-/// Build the queue of page numbers `TargetSource::discover` will walk.
+/// Walker returned by [`TargetSource::discover`]. Walks pages `1..` in
-/// The site orders by `update_date DESC`, so newest-first is just the
+/// order, terminating as soon as a page renders cleanly with zero entries
-/// natural page order: `1..=last`. If `last_page` is unknown (source
+/// — that's the "we ran off the end of the index" signal. Page 1's HTML
-/// surfaces no pagination) only page 1 is visited.
+/// is cached at construction time (discover already had to fetch it for
-fn build_page_order(last_page: Option<i32>) -> VecDeque<i32> {
+/// the transient probe) so the first batch doesn't re-fetch.
-    match last_page {
+///
-        None => VecDeque::from([1]),
+/// A genuinely empty `Ok(vec![])` from `parse_manga_list_from` is what
-        Some(last) => (1..=last).collect(),
+/// stops us: the parser's `#logo` sentinel converts unrendered pages
-    }
+/// into transient errors before they reach this loop, so an empty
-}
+/// parse result reliably means "no more entries."
-
-/// Walker returned by [`TargetSource::discover`]. Pops one source-index
-/// page per `next_batch` call. Page 1's HTML is cached at construction
-/// time (the discover call needed it to read `last_page` anyway) so the
-/// batch covering page 1 doesn't re-fetch.
 struct TargetSourceWalker {
     base_url: String,
-    pages_remaining: VecDeque<i32>,
+    next_page: i32,
     first_page_html: Option<String>,
 }
 
@@ -174,13 +158,11 @@ impl DiscoverWalk for TargetSourceWalker {
         &mut self,
         ctx: &FetchContext<'_>,
     ) -> anyhow::Result<Option<Vec<SourceMangaRef>>> {
-        let Some(page_num) = self.pages_remaining.pop_front() else {
+        let page_num = self.next_page;
-            return Ok(None);
-        };
         let page_refs = if page_num == 1 {
             // Reuse the cached page-1 HTML from the initial probe. Take
-            // it (rather than clone) so a malformed page-order queue
+            // it (rather than clone) so a future re-entry that somehow
-            // that re-visits page 1 still falls back to a real fetch.
+            // revisits page 1 still falls back to a real fetch.
             match self.first_page_html.take() {
                 Some(html) => {
                     let doc = scraper::Html::parse_document(&html);
@@ -218,6 +200,10 @@ impl DiscoverWalk for TargetSourceWalker {
             .await?
         };
         tracing::info!(page_num, count = page_refs.len(), "page walked");
+        if page_refs.is_empty() {
+            return Ok(None);
+        }
+        self.next_page += 1;
         Ok(Some(page_refs))
     }
 }
@@ -288,20 +274,6 @@ fn classify_navigate_html(html: String) -> Result<String, PageError> {
     Ok(html)
 }
 
-fn parse_last_page(doc: &scraper::Html) -> Option<i32> {
-    // Pagination links carry their page number as text. Take the
-    // numeric maximum so we don't depend on a specific layout (Prev,
-    // Next, ellipses, etc. all get filtered out by .parse).
-    let sel = scraper::Selector::parse("#left_side .pagination a").unwrap();
-    doc.select(&sel)
-        .filter_map(|a| {
-            collapse_whitespace(&a.text().collect::<String>())
-                .parse::<i32>()
-                .ok()
-        })
-        .max()
-}
-
 /// Substitutes the first `/N/` path segment with the target page
 /// number. Source impls that paginate via a different URL shape can
 /// override this — for the modeled site the segment is always present.
@@ -853,29 +825,6 @@ mod tests {
         assert_eq!(parse_chapter_number("Special"), None);
     }
 
-    #[test]
-    fn parse_last_page_picks_highest_pagination_link() {
-        let html = r#"
-            <div id="left_side"><div class="pagination">
-              <a href="/list/1/">Prev</a>
-              <ol>
-                <li><a href="/list/1/">1</a></li>
-                <li><a href="/list/2/">2</a></li>
-                <li><a href="/list/47/">47</a></li>
-                <li><a href="/list/2/">Next</a></li>
-              </ol>
-            </div></div>
-        "#;
-        let doc = scraper::Html::parse_document(html);
-        assert_eq!(parse_last_page(&doc), Some(47));
-    }
-
-    #[test]
-    fn parse_last_page_none_when_no_pagination() {
-        let doc = scraper::Html::parse_document("<html></html>");
-        assert!(parse_last_page(&doc).is_none());
-    }
-
     #[test]
     fn page_url_substitutes_numeric_path_segment() {
         assert_eq!(
@@ -1024,28 +973,6 @@ mod tests {
         assert!(err.is_transient(), "got non-transient: {err}");
     }
 
-    #[test]
-    fn build_page_order_is_natural_one_to_last() {
-        // Newest-first is just the source's natural pagination order:
-        // (update_date DESC) lives at page 1, oldest at the last page.
-        let order = build_page_order(Some(3));
-        assert_eq!(Vec::from(order), vec![1, 2, 3]);
-    }
-
-    #[test]
-    fn build_page_order_falls_back_to_page_one_only_without_pagination() {
-        // Source surfaced no pagination control — visit page 1 alone
-        // and let the walk end after one batch.
-        let order = build_page_order(None);
-        assert_eq!(Vec::from(order), vec![1]);
-    }
-
-    #[test]
-    fn build_page_order_single_page_index_yields_one_entry() {
-        let order = build_page_order(Some(1));
-        assert_eq!(Vec::from(order), vec![1]);
-    }
-
     #[test]
     fn parse_chapter_list_returns_transient_when_table_missing() {
         // Partial render (post-load JS hadn't injected the table, layout

frontend/package.json

@@ -1,6 +1,6 @@
 {
   "name": "mangalord-frontend",
-  "version": "0.45.0",
+  "version": "0.45.1",
   "private": true,
   "type": "module",
   "scripts": {