fix(web): show TAR hard links in archive previews

apps/web/src/components/files/AttachmentFilePreview.browser.test.tsx

@@ -85,13 +85,20 @@ import type { OfficeWorkerResult } from "./officePreviewProtocol";
 describe("bounded binary previews", () => {
   it("preserves length-delimited local and global PAX attributes", async () => {
     const encode = (value: string) => new TextEncoder().encode(value);
-    const block = (name: string, content: Uint8Array, type = "0", size = content.length) => {
+    const block = (
+      name: string,
+      content: Uint8Array,
+      type = "0",
+      size = content.length,
+      linkname = "",
+    ) => {
       const bytes = new Uint8Array(512 + Math.ceil(content.length / 512) * 512);
       const write = (offset: number, value: string) => bytes.set(encode(value), offset);
       write(0, name);
       write(124, size.toString(8).padStart(11, "0") + "\0");
       bytes.fill(32, 148, 156);
       write(156, type);
+      write(157, linkname);
       write(
         148,
         bytes
@@ -123,6 +130,15 @@ describe("bounded binary previews", () => {
       block("Local", attributes({ size: "3", mtime: "-172800" }), "x"),
       block("short.txt", encode("bye"), "0", 0),
       block("again.txt", encode("again"), "0", 0),
+      block("Global", attributes({ size: "", mtime: "" }), "g"),
+      block("original.txt", encode("hello world")),
+      block("alias.txt", encode(""), "1", 0, "original.txt"),
+      block("chain.txt", encode(""), "1", 0, "alias.txt"),
+      block("Local", attributes({ linkpath: "original.txt" }), "x"),
+      block("pax-link.txt", encode(""), "1", 0, "wrong-target"),
+      block("LongLink", encode("original.txt\0"), "K"),
+      block("gnu-link.txt", encode(""), "1", 0, "wrong-target"),
+      block("original.txt", encode("replacement")),
       new Uint8Array(1024),
     ];
     const bytes = new Uint8Array(chunks.reduce((size, chunk) => size + chunk.length, 0));
@@ -154,6 +170,14 @@ describe("bounded binary previews", () => {
         ["world.txt", 5, "1969-12-31T00:00:00.000Z"],
         ["short.txt", 3, "1969-12-30T00:00:00.000Z"],
         ["again.txt", 5, "1969-12-31T00:00:00.000Z"],
+        ...[
+          "original.txt",
+          "alias.txt",
+          "chain.txt",
+          "pax-link.txt",
+          "gnu-link.txt",
+          "original.txt",
+        ].map((path) => [path, 11, "1970-01-01T00:00:00.000Z"]),
       ]);
       const entry = await request({ type: "entry", path: "café\n9 size=1\nname.txt" });
       if (entry.type !== "entry") throw new Error("Expected entry");
@@ -164,6 +188,42 @@ describe("bounded binary previews", () => {
       if (reopened.type !== "entry") throw new Error("Expected entry");
       expect(new TextDecoder().decode(reopened.bytes)).toBe("hello");
       expect(reopened.bytes.buffer.byteLength).toBe(5);
+      for (const path of ["alias.txt", "chain.txt", "pax-link.txt", "gnu-link.txt"]) {
+        const linked = await request({ type: "entry", path });
+        if (linked.type !== "entry") throw new Error("Expected hard link entry");
+        expect(new TextDecoder().decode(linked.bytes)).toBe("hello world");
+      }
+      const unresolved = new Uint8Array(
+        await new Blob([
+          block("missing.txt", encode(""), "1", 0, "nowhere.txt"),
+          block("early.txt", encode(""), "1", 0, "later.txt"),
+          block("later.txt", encode("hello")),
+          block("folder/", encode(""), "5"),
+          block("to-folder.txt", encode(""), "1", 0, "folder/"),
+          block("replaced.txt", encode("stale")),
+          block("replaced.txt", encode(""), "2", 0, "elsewhere.txt"),
+          block("to-replaced.txt", encode(""), "1", 0, "replaced.txt"),
+          block("Local", attributes({ linkpath: "" }), "x"),
+          block("emptied.txt", encode(""), "1", 0, "later.txt"),
+          new Uint8Array(1024),
+        ]).arrayBuffer(),
+      );
+      const skipped = await request({
+        type: "open",
+        kind: "archive",
+        name: "links.tar",
+        bytes: unresolved.buffer,
+      });
+      if (skipped.type !== "archive") throw new Error("Expected archive");
+      expect(skipped.entries.map((entry) => entry.path)).toEqual([
+        "later.txt",
+        "folder/",
+        "replaced.txt",
+      ]);
+      expect(await request({ type: "entry", path: "replaced.txt" })).toEqual({
+        type: "error",
+        message: "This archive entry is unavailable.",
+      });
     } finally {
       worker.terminate();
     }

apps/web/src/components/files/binaryPreview.worker.ts

@@ -52,6 +52,7 @@ function tarEntries(bytes: Uint8Array<ArrayBuffer>) {
     return Number.parseInt(value || "0", 8);
   };
   let longPath: string | null = null;
+  let longLink: string | null = null;
   const globalPax: Record<string, string> = {};
   let localPax: Record<string, string> = {};
   let offset = 0;
@@ -64,7 +65,7 @@ function tarEntries(bytes: Uint8Array<ArrayBuffer>) {
       checksum += index >= 148 && index < 156 ? 32 : bytes[offset + index]!;
     if (checksum !== number(offset + 148, 8)) throw new Error("The TAR header is damaged.");
     const type = text(offset + 156, 1);
-    const metadata = type === "L" || type === "x" || type === "g";
+    const metadata = type === "L" || type === "K" || type === "x" || type === "g";
     const attributes: Record<string, string> = metadata ? {} : { ...globalPax, ...localPax };
     const size = attributes.size ? Number(attributes.size) : number(offset + 124, 12);
     if (attributes.size && !/^\d+$/.test(attributes.size))
@@ -75,7 +76,9 @@ function tarEntries(bytes: Uint8Array<ArrayBuffer>) {
         : "";
     const path =
       attributes.path || longPath || [prefix, text(offset, 100)].filter(Boolean).join("/");
+    const linkpath = attributes.linkpath ?? longLink ?? text(offset + 157, 100);
     if (!metadata) {
+      longLink = null;
       longPath = null;
       localPax = {};
     }
@@ -89,6 +92,7 @@ function tarEntries(bytes: Uint8Array<ArrayBuffer>) {
       throw new Error("An archive entry is too large or damaged.");
     const content = bytes.subarray(offset + 512, offset + 512 + size);
     if (type === "L") longPath = new TextDecoder().decode(content).replace(/\0.*$/s, "");
+    else if (type === "K") longLink = new TextDecoder().decode(content).replace(/\0.*$/s, "");
     else if (type === "x" || type === "g") {
       if (size > 65536) throw new Error("Archive metadata is too large.");
       for (let cursor = 0; cursor < content.length;) {
@@ -111,26 +115,38 @@ function tarEntries(bytes: Uint8Array<ArrayBuffer>) {
         if (equals < 1) throw new Error("The TAR metadata is damaged.");
         const key = record.slice(0, equals);
         const value = record.slice(equals + 1);
-        if (["path", "size", "mtime"].includes(key)) {
+        if (["path", "linkpath", "size", "mtime"].includes(key)) {
           if (type === "x") localPax[key] = value;
-          else if (value) globalPax[key] = value;
+          else if (value || key === "linkpath") globalPax[key] = value;
           else delete globalPax[key];
         }
         cursor = end;
       }
-    } else if (type === "" || type === "0" || type === "5") {
+    } else if (
+      type === "" ||
+      type === "0" ||
+      type === "5" ||
+      // A hard link shares an earlier member's bytes; one without that member is skipped.
+      (type === "1" && linkpath && files.has(linkpath))
+    ) {
       if (attributes.mtime && !/^[+-]?\d+(?:\.\d+)?$/.test(attributes.mtime))
         throw new Error("The TAR metadata is damaged.");
       const modified =
         (attributes.mtime ? Number(attributes.mtime) : number(offset + 136, 12)) * 1000;
       const date = new Date(modified);
+      const linked = type === "1" ? files.get(linkpath) : undefined;
       entries.push({
         path,
-        size,
+        size: linked?.length ?? size,
         directory: type === "5",
         modified: Number.isNaN(date.getTime()) ? null : date.toISOString(),
       });
-      if (type !== "5") files.set(path, content.slice());
+      if (type === "5") {
+        files.delete(path);
+        files.delete(path.replace(/\/+$/, ""));
+      } else files.set(path, linked ?? content.slice());
+    } else {
+      files.delete(path);
     }
     offset += 512 + Math.ceil(size / 512) * 512;
   }