Add test cases for redirected responses

[Rob--W]

These are regression tests for issue #12744 and PR #19028.

[Rob--W]

Note: rebase this when #19028 is merged. The new test is expected to fail right now because the fix has not been applied yet.

[Snuffleupagus]

Thanks for helping out with the tests!

[Rob--W]

@Snuffleupagus I have addressed the nits, and also updated the PR to run the tests for all IPDFStream implementations where this is relevant, instead of just PDFNetworkStream - i.e. I repeated the test for PDFFetchStream.

[Rob--W]

@Snuffleupagus I've updated the test by moving it to a separate file. After doing that, I ran into a new issue: while the network_spec.js test ran in the browser only, the fetch_stream_spec.js test was also run in Node.js. I looked into re-using the existing "temporary node server" helper, but then I would have to re-implement the redirect functionality of webserver.mjs there.

So in the end I decided to prepend another commit that replaces createTemporaryNodeServer with a new test helper that offers a uniform test server interface. Please let me know what you think of it. I can also split that test server to a separate PR if you'd like.

[Snuffleupagus]

Locally (on Windows) the new tests unfortunately seem to fail when running gulp unittest, which is the "normal" way to run the unit-tests and also what's used on the bots.

/botio unittest

[Rob--W]

Locally (on Windows) the new tests unfortunately seem to fail when running gulp unittest, which is the "normal" way to run the unit-tests and also what's used on the bots.

Sorry about that, I accidentally lost some lines while rebasing.

This is the diff compared to before:

diff --git a/test/unit/network_spec.js b/test/unit/network_spec.js
index 4c2624a8b..367e52560 100644
--- a/test/unit/network_spec.js
+++ b/test/unit/network_spec.js
@@ -16,6 +16,7 @@
 import { AbortException } from "../../src/shared/util.js";
 import { PDFNetworkStream } from "../../src/display/network.js";
 import { testCrossOriginRedirects } from "./common_pdfstream_tests.js";
+import { TestPdfsServer } from "./test_utils.js";
 
 describe("network", function () {
   const pdf1 = new URL("../pdfs/tracemonkey.pdf", window.location).href;
@@ -118,6 +119,14 @@ describe("network", function () {
   });
 
   describe("Redirects", function () {
+    beforeAll(async function () {
+      await TestPdfsServer.ensureStarted();
+    });
+
+    afterAll(async function () {
+      await TestPdfsServer.ensureStopped();
+    });
+
     it("redirects allowed if all responses are same-origin", async function () {
       await testCrossOriginRedirects({
         PDFStreamClass: PDFNetworkStream,

[Rob--W]

Oddly, the test passes in Firefox but failed in Chrome, with the following error:

TEST-UNEXPECTED-FAIL | redirects blocked if any response is cross-origin | in chrome | Expected a promise to be rejected but it was resolved. Expected a promise to be rejected but it was resolved.

I updated the patch to add a cache buster, to see if that fixes the test.

/botio unittest

[moz-tools-bot]

From: Bot.io (Linux m4)

---

Failed

Full output at http://54.241.84.105:8877/efefda0bf24c87e/output.txt

Total script time: 2.48 mins

• Unit Tests: FAILED

[Rob--W]

??? The test output is inconsistent. Within the same file in the test on Linux, I see the following (one for network, one for fetch_stream, unclear which is which): http://54.241.84.105:8877/efefda0bf24c87e/output.txt

TEST-UNEXPECTED-FAIL | redirects blocked if any response is cross-origin | in chrome | Expected a promise to be rejected but it was resolved. Expected a promise to be rejected but it was resolved.
TEST-PASSED | redirects blocked if any response is cross-origin | in chrome

[Rob--W]

Another attempt - adding Vary:Origin at the server side instead of cache buster at the client side.

[Rob--W]

/botio unittest

[moz-tools-bot]

From: Bot.io (Windows)

---

Received

Command cmd_unittest from @Rob--W received. Current queue size: 0

Live output at: http://54.193.163.58:8877/cf821e7cebb656b/output.txt

[moz-tools-bot]

From: Bot.io (Linux m4)

---

Received

Command cmd_unittest from @Rob--W received. Current queue size: 0

Live output at: http://54.241.84.105:8877/203b9ff7350ccec/output.txt

[moz-tools-bot]

From: Bot.io (Linux m4)

---

Failed

Full output at http://54.241.84.105:8877/203b9ff7350ccec/output.txt

Total script time: 2.53 mins

• Unit Tests: FAILED

[moz-tools-bot]

From: Bot.io (Windows)

---

Failed

Full output at http://54.193.163.58:8877/cf821e7cebb656b/output.txt

Total script time: 6.44 mins

• Unit Tests: FAILED

[Rob--W]

I suspect that Chrome serves the range response directly from its own cache (based on the response from the full request), without reaching the server, and thus never triggering the redirect, and therefore not triggering the expected failure.

I'll look into a way to avoid that, later.

[Rob--W]

I've added Cache-Control: no-store,max-age=0 to the relevant requests from the new test.

/botio unittest

[moz-tools-bot]

From: Bot.io (Linux m4)

---

Received

Command cmd_unittest from @Rob--W received. Current queue size: 0

Live output at: http://54.241.84.105:8877/3bf4c156cad42c5/output.txt

[moz-tools-bot]

From: Bot.io (Windows)

---

Received

Command cmd_unittest from @Rob--W received. Current queue size: 1

Live output at: http://54.193.163.58:8877/180e7f620d54e71/output.txt

[moz-tools-bot]

From: Bot.io (Linux m4)

---

Success

Full output at http://54.241.84.105:8877/3bf4c156cad42c5/output.txt

Total script time: 2.49 mins

• Unit Tests: Passed

[moz-tools-bot]

From: Bot.io (Windows)

---

Success

Full output at http://54.193.163.58:8877/180e7f620d54e71/output.txt

Total script time: 6.36 mins

• Unit Tests: Passed

[Rob--W]

@Snuffleupagus The tests now pass in Chrome, it looks like my analysis of the issue along with the fix is correct. Please review again.

Analysis:

I suspect that Chrome serves the range response directly from its own cache (based on the response from the full request), without reaching the server, and thus never triggering the redirect, and therefore not triggering the expected failure.

I'll look into a way to avoid that, later.

Fix:

I've added Cache-Control: no-store,max-age=0 to the relevant requests from the new test.

[Snuffleupagus]

r=me, assuming passing unit-tests; thank you.

[Snuffleupagus]

/botio unittest

[moz-tools-bot]

From: Bot.io (Linux m4)

---

Received

Command cmd_unittest from @Snuffleupagus received. Current queue size: 0

Live output at: http://54.241.84.105:8877/6ab44c041e86752/output.txt

[moz-tools-bot]

From: Bot.io (Windows)

---

Received

Command cmd_unittest from @Snuffleupagus received. Current queue size: 0

Live output at: http://54.193.163.58:8877/37771b586565f75/output.txt

[moz-tools-bot]

From: Bot.io (Linux m4)

---

Success

Full output at http://54.241.84.105:8877/6ab44c041e86752/output.txt

Total script time: 2.50 mins

• Unit Tests: Passed

[moz-tools-bot]

From: Bot.io (Windows)

---

Success

Full output at http://54.193.163.58:8877/37771b586565f75/output.txt

Total script time: 6.77 mins

• Unit Tests: Passed