It looks good to me!

I opened #24442 with this patch. But we really should write a test too.

Do either of you know how to re-produce the issue and make a test?

I know how to re-produce it in principle (simply by trying to read from the file past the end of it). However I can't get the local tests to work (yet), they are all failing, so I can't write a test right now:

EMTEST_BROWSER=chromium ./test/runner.py browser.test_wasmfs_fetch*

--- expected
+++ actual
@@ -1 +1 @@
-/report_result?exit:0
+/report_result?abort:Assertion failed: read_now > 0, at: /home/foxtacles/emsdk/upstream/emscripten/test/wasmfs/wasmfs_fetch.c,136,read_chunks_check

Probably missing something obvious since it's my first time running the tests. Other tests seem to pass.

I've managed to get the tests running using Firefox.

Unfortunately, reproducing this issue with tests turned out much harder for 2 reasons:

1. Using the read API, I somehow can't manage to nudge the file offset past the end of the file (which actually seems like sound behavior). I'm not sure how it was possible that this happened in the first place in my Emscripten app (but it definitely happens and I can still reproduce it there). This might hint to another bug somewhere else.
2. Whenever I set the offset manually past the end of the file (subsequent Range request has a start offset past the file size), I'm unable to reproduce the problem because the test server the request is issued against never returns 416 Range Not Satisfiable, but instead returns a 2xx with an empty body. That seems incorrect.

If you are referring to test server using by the emscripten test suite that seems like something we should be able to fix (re generating 416 error).

The fact that you cannot make it happen using normal read calls is more worrying. If you are able to reproduce in your full app can you capture a full backtrace so we can see where call comes from.