browser.test_cross_origin regression: "Expected to find '/report_result?exit:0 ' in '/report_result?abort:Assertion failed"

[juj]

After PR #27018

commit 41656d690b521208189ed838c5ab474435b208c9
Author: Sam Clegg <sbc@chromium.org>
Date:   Thu May 28 14:29:24 2026 -0700

    [pthreads] Simplify message relaying code (#27018)

    Historically, `targetThread` relaying was used to pass `objectTransfer`
    messages (containing `OffscreenCanvas` objects) back to parent threads
    on worker exit. This was introduced in commit 1b278917cd (2018/2019) and
    required supporting `transferList` in the relaying code.

    However, automatic canvas return on exit was later disabled in
    6dcf7298e8 (2019) and the `objectTransfer` command was fully removed in
    0f536c60bb (2021).

    These days, the relaying mechanism is only used by `CMD_CHECK_MAILBOX`
    (via `_emscripten_notify_mailbox_postmessage`) to send simple mailbox
    notifications, which do not require a transfer list.

the test browser.test_cross_origin (and possibly several other tests, as the Firefox suites are now very red) are failing with

C:\emsdk\emscripten\main>test\runner browser.test_cross_origin
Running 1 tests
[1/1] test_cross_origin (test_browser.browser.test_cross_origin) ... FAIL

Stderr:
127.0.0.1 - - [31/May/2026 20:27:49] code 404, message File not found
127.0.0.1 - - [31/May/2026 20:27:49] "GET /hello.js HTTP/1.1" 200 -
127.0.0.1 - - [31/May/2026 20:27:49] "GET /hello.wasm HTTP/1.1" 200 -
127.0.0.1 - - [31/May/2026 20:27:49] "GET /hello.js HTTP/1.1" 200 -


======================================================================
FAIL: test_cross_origin (test_browser.browser.test_cross_origin)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "C:\emsdk\emscripten\main\test\common.py", line 380, in resulting_test
    return func(self, *args)
  File "C:\emsdk\emscripten\main\test\test_browser.py", line 5659, in test_cross_origin
    self.run_browser('test.html', '/report_result?exit:0')
    ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\emsdk\emscripten\main\test\browser_common.py", line 899, in run_browser
    raise e
  File "C:\emsdk\emscripten\main\test\browser_common.py", line 890, in run_browser
    self.assertContained(expected, output)
    ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^
  File "C:\emsdk\emscripten\main\test\common.py", line 1136, in assertContained
    self.fail("Expected to find '%s' in '%s', diff:\n\n%s\n%s" % (
    ~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      limit_size(values[0]), limit_size(string), limit_size(diff),
      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
      additional_info,
      ^^^^^^^^^^^^^^^^
    ))
    ^^
AssertionError: Expected to find '/report_result?exit:0
' in '/report_result?abort:Assertion failed
', diff:

--- expected
+++ actual
@@ -1 +1 @@
-/report_result?exit:0
+/report_result?abort:Assertion failed



Stderr:
127.0.0.1 - - [31/May/2026 20:27:49] code 404, message File not found
127.0.0.1 - - [31/May/2026 20:27:49] "GET /hello.js HTTP/1.1" 200 -
127.0.0.1 - - [31/May/2026 20:27:49] "GET /hello.wasm HTTP/1.1" 200 -
127.0.0.1 - - [31/May/2026 20:27:49] "GET /hello.js HTTP/1.1" 200 -

----------------------------------------------------------------------
Ran 1 test in 2.475s

FAILED (failures=1)

C:\emsdk\emscripten\main>

[juj]

Quick look gives this:

[juj]

The test sockets.test_posix_proxy_sockets fails by the same commit, and it gives

where test.js:1476 is

            // pthreads should not be relaying messages to themselves.
            assert(d.targetThread != _pthread_self());

In the test, d.targetThread == _pthread_self() == 71748

[sbc100]

Wow, very strange. Is this just effecting firefox, right?

The first screenshot you posted looks like the test succeeds (the error there is just regarding sending the exit:0 result back to the test hardness which runs on a different port, I think thats a success).

Regarding the assert(d.targetThread != _pthread_self()); , maybe I'm misunderstanding because I don't see how that could ever happen.

IIUC the only user of the targetThread field is _emscripten_notify_mailbox_postmessage and it has a special case there that means you can never sent this message to yourself.

Did you miss some other usage of targetThread in the codebase somehow ?

[juj]

Wow, very strange. Is this just effecting firefox, right?

Looks like so. Safari runs on macOS 26 at http://clbri.com:8010/#/builders/13/builds/419 are ok: http://clbri.com:8010/api/v2/logs/467502/raw_inline . I don't know of Chrome, since my CI doesn't have it.

The first screenshot you posted looks like the test succeeds (the error there is just regarding sending the exit:0 result back to the test hardness which runs on a different port, I think thats a success).

That probably was because I ran the test.html with emrun from out/test/, so it used a different port.

Regarding the assert(d.targetThread != _pthread_self()); , maybe I'm misunderstanding because I don't see how that could ever happen.

IIUC the only user of the targetThread field is _emscripten_notify_mailbox_postmessage and it has a special case there that means you can never sent this message to yourself.

Did you miss some other usage of targetThread in the codebase somehow ?

I did not modify the code here, this is all on the CI on upstream code.

[sbc100]

I did not modify the code here, this is all on the CI on upstream code.

Sorry, I didn't mean to imply that. I was looking that #27018 (the change you bisected to) and trying to see how it could have possibly had these effects. We can revert it of course, but I'd love to understand how/why that change could have caused these kind of issues.