Merge pull request #15613 from Snuffleupagus/issue-15590

[api-minor] Let `Catalog.getAllPageDicts` return an *empty*  dictionary when loading the first /Page fails (issue 15590)

src/core/catalog.js

@@ -1191,6 +1191,8 @@ class Catalog {
    * @returns {Promise<Map>}
    */
   async getAllPageDicts(recoveryMode = false) {
+    const { ignoreErrors } = this.pdfManager.evaluatorOptions;
+
     const queue = [{ currentNode: this.toplevelPagesDict, posInKids: 0 }];
     const visitedNodes = new RefSet();
 
@@ -1215,6 +1217,11 @@ class Catalog {
       if (error instanceof XRefEntryException && !recoveryMode) {
         throw error;
       }
+      if (recoveryMode && ignoreErrors && pageIndex === 0) {
+        // Ensure that the viewer will always load (fixes issue15590.pdf).
+        warn(`getAllPageDicts - Skipping invalid first page: "${error}".`);
+        error = Dict.empty;
+      }
 
       map.set(pageIndex++, [error, null]);
     }

src/core/xref.js

@@ -503,7 +503,7 @@ class XRef {
         // Find the next "obj" string, rather than "endobj", to ensure that
         // we won't skip over a new 'obj' operator in corrupt files where
         // 'endobj' operators are missing (fixes issue9105_reduced.pdf).
-        while (startPos < buffer.length) {
+        while (startPos < length) {
           const endPos = startPos + skipUntil(buffer, startPos, objBytes) + 4;
           contentLength = endPos - position;
 
@@ -545,7 +545,29 @@ class XRef {
         (token.length === 7 || /\s/.test(token[7]))
       ) {
         trailers.push(position);
-        position += skipUntil(buffer, position, startxrefBytes);
+
+        const contentLength = skipUntil(buffer, position, startxrefBytes);
+        // Attempt to handle (some) corrupt documents, where no 'startxref'
+        // operators are present (fixes issue15590.pdf).
+        if (position + contentLength >= length) {
+          const endPos = position + skipUntil(buffer, position, objBytes) + 4;
+
+          const checkPos = Math.max(endPos - CHECK_CONTENT_LENGTH, position);
+          const tokenStr = bytesToString(buffer.subarray(checkPos, endPos));
+
+          // Find the first "obj" occurrence after the 'trailer' operator.
+          const objToken = nestedObjRegExp.exec(tokenStr);
+
+          if (objToken && objToken[1]) {
+            warn(
+              'indexObjects: Found first "obj" after "trailer", ' +
+                'caused by missing "startxref" -- trying to recover.'
+            );
+            position = endPos - objToken[1].length;
+            continue;
+          }
+        }
+        position += contentLength;
       } else {
         position += token.length + 1;
       }

test/pdfs/.gitignore

@@ -351,6 +351,7 @@
 !issue9534_reduced.pdf
 !attachment.pdf
 !basicapi.pdf
+!issue15590.pdf
 !issue15594_reduced.pdf
 !issue2884_reduced.pdf
 !mixedfonts.pdf

test/pdfs/issue15590.pdf

@@ -0,0 +1,23 @@
+%PDF-1.7
+
+trailer
+<<
+/Root 1 0 R
+>>
+
+1 0 obj
+<<
+/Type /Catalog
+/Pages 2 0 R
+/OpenAction 2 0 R
+>>
+endobj
+
+2 0 obj
+<<
+/S /JavaScript
+/JS(func=function(){app.alert(1)};func();)
+>>
+endobj
+
+%%EOF

test/unit/api_spec.js

@@ -602,27 +602,42 @@ describe("api", function () {
       const loadingTask2 = getDocument(
         buildGetDocumentParams("poppler-85140-0.pdf")
       );
+      const loadingTask3 = getDocument(
+        buildGetDocumentParams("poppler-85140-0.pdf", { stopAtErrors: true })
+      );
 
       expect(loadingTask1 instanceof PDFDocumentLoadingTask).toEqual(true);
       expect(loadingTask2 instanceof PDFDocumentLoadingTask).toEqual(true);
+      expect(loadingTask3 instanceof PDFDocumentLoadingTask).toEqual(true);
 
       const pdfDocument1 = await loadingTask1.promise;
       const pdfDocument2 = await loadingTask2.promise;
+      const pdfDocument3 = await loadingTask3.promise;
 
       expect(pdfDocument1.numPages).toEqual(1);
       expect(pdfDocument2.numPages).toEqual(1);
+      expect(pdfDocument3.numPages).toEqual(1);
 
-      const page = await pdfDocument1.getPage(1);
+      const pageA = await pdfDocument1.getPage(1);
-      expect(page instanceof PDFPageProxy).toEqual(true);
+      expect(pageA instanceof PDFPageProxy).toEqual(true);
 
-      const opList = await page.getOperatorList();
+      const opListA = await pageA.getOperatorList();
-      expect(opList.fnArray.length).toBeGreaterThan(5);
+      expect(opListA.fnArray.length).toBeGreaterThan(5);
-      expect(opList.argsArray.length).toBeGreaterThan(5);
+      expect(opListA.argsArray.length).toBeGreaterThan(5);
-      expect(opList.lastChunk).toEqual(true);
+      expect(opListA.lastChunk).toEqual(true);
-      expect(opList.separateAnnots).toEqual(null);
+      expect(opListA.separateAnnots).toEqual(null);
+
+      const pageB = await pdfDocument2.getPage(1);
+      expect(pageB instanceof PDFPageProxy).toEqual(true);
+
+      const opListB = await pageB.getOperatorList();
+      expect(opListB.fnArray.length).toBe(0);
+      expect(opListB.argsArray.length).toBe(0);
+      expect(opListB.lastChunk).toEqual(true);
+      expect(opListB.separateAnnots).toEqual(null);
 
       try {
-        await pdfDocument2.getPage(1);
+        await pdfDocument3.getPage(1);
 
         // Shouldn't get here.
         expect(false).toEqual(true);
@@ -631,7 +646,11 @@ describe("api", function () {
         expect(reason.message).toEqual("Bad (uncompressed) XRef entry: 3R");
       }
 
-      await Promise.all([loadingTask1.destroy(), loadingTask2.destroy()]);
+      await Promise.all([
+        loadingTask1.destroy(),
+        loadingTask2.destroy(),
+        loadingTask3.destroy(),
+      ]);
     });
 
     it("creates pdf doc from PDF files, with circular references", async function () {
@@ -738,6 +757,24 @@ describe("api", function () {
 
       await loadingTask.destroy();
     });
+
+    it("creates pdf doc from PDF file, with incomplete trailer", async function () {
+      const loadingTask = getDocument(buildGetDocumentParams("issue15590.pdf"));
+      expect(loadingTask instanceof PDFDocumentLoadingTask).toEqual(true);
+
+      const pdfDocument = await loadingTask.promise;
+      expect(pdfDocument.numPages).toEqual(1);
+
+      const jsActions = await pdfDocument.getJSActions();
+      expect(jsActions).toEqual({
+        OpenAction: ["func=function(){app.alert(1)};func();"],
+      });
+
+      const page = await pdfDocument.getPage(1);
+      expect(page instanceof PDFPageProxy).toEqual(true);
+
+      await loadingTask.destroy();
+    });
   });
 
   describe("PDFWorker", function () {