Merge pull request #925 from tdurieux/fix/unavailable-submodule-commit

fix: keep repositories readable with invalid submodule commits

src/core/source/github-submodules.ts

@@ -58,7 +58,17 @@ export async function expandSubmodules(files: IFile[], data: GitHubBaseData,
       await waitForTokenGate("");
       const child = await octokit("").git.getTree({
         owner: repository.organization, repo: repository.repoName, tree_sha: mount.sha, recursive: "1",
+      }).catch((error: unknown) => {
+        const upstream = error as { status?: number; response?: { data?: { message?: string } } };
+        // GitHub also uses 422 for a gitlink whose pinned object is missing.
+        // Only classify this child-tree response, not arbitrary validation errors.
+        if (upstream.status === 422 && /^Invalid object requested\. SHA must identify a commit or a tree\b/.test(upstream.response?.data?.message || "")) {
+          mount.submodule = "unavailable";
+          return null;
+        }
+        throw error;
       });
+      if (!child) continue;
       if (child.data.truncated || child.data.tree.length > budget.remaining) { mount.submodule = "limit"; continue; }
       const entries = child.data.tree;
       if (entries.some(entry => !safeRepositoryPath(entry.path))) { mount.submodule = "unsupported"; continue; }

test/github-submodules.test.js

@@ -138,6 +138,40 @@ describe("public GitHub submodules", function () {
     expect(calls).to.have.length(1);
   });
 
+  it("keeps parent files when a child gitlink points to an invalid Git object", async () => {
+    const error = Object.assign(new Error("invalid object"), { status: 422, response: { data: {
+      message: "Invalid object requested. SHA must identify a commit or a tree.",
+    } } });
+    api([], { error });
+    const source = new GitHubStream(parent);
+    source.getGHTree = async () => ({ tree: [
+      { path: "vendor/dependency", type: "commit", mode: "160000", sha },
+      { path: "README.md", type: "blob", sha: blob, size: 5 },
+    ] });
+    const files = await source.getFiles();
+    expect(files.find(file => file.name === "dependency").submodule).to.equal("unavailable");
+    expect(files.find(file => file.name === "dependency").upstream).not.to.exist;
+    expect(files.find(file => file.name === "README.md").sha).to.equal(blob);
+  });
+
+  it("does not suppress invalid-object errors from parent metadata", async () => {
+    const error = Object.assign(new Error("invalid parent"), { status: 422, response: { data: {
+      message: "Invalid object requested. SHA must identify a commit or a tree.",
+    } } });
+    stub(utils, "octokit", () => ({ repos: { getContent: async () => { throw error; } } }));
+    let caught; try { await expandSubmodules([mount()], parent); } catch (err) { caught = err; }
+    expect(caught).to.equal(error);
+  });
+
+  for (const [status, message] of [[422, "Validation Failed"], [422, ""], [429, "rate limited"], [500, "server error"]]) {
+    it(`preserves child failure ${status}: ${message}`, async () => {
+      const error = Object.assign(new Error(message), { status, response: { data: { message } } });
+      api([], { error });
+      let caught; try { await expandSubmodules([mount()], parent); } catch (err) { caught = err; }
+      expect(caught).to.equal(error);
+    });
+  }
+
   it("does not persist transient failures as permanently missing dependencies", async () => {
     const error = Object.assign(new Error("rate limited"), { status: 403 }); api([], { error });
     let caught; try { await expandSubmodules([mount()], parent); } catch (err) { caught = err; }