1687903 - Consider returning a network error if status-code cannot be parsed

Status: RESOLVED FIXED

(Core :: Networking: HTTP, enhancement, P3)

Description

[Anne (:annevk)]

Currently this exhibits rather unusual behavior, especially for numbers beyond 65535. Safari already does this for numbers with more than three digits.

https://github.com/whatwg/fetch/issues/1142

Comment 1

[Anne (:annevk)]

I have some tests at https://github.com/web-platform-tests/wpt/pull/27421. Safari might be stricter due to not supporting HTTP/0.9 anymore?

Comment 2

[Thomas Wisniewski [:twisniewski]]

Attachment: Bug 1687903 - Enable stricter HTTP response status line parsing; r?valentin

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/4e21d241af64
Enable stricter HTTP response status line parsing; r=valentin,necko-reviewers

Comment 4

[Natalia Csoregi [:nataliaCs]]

Backed out for causing multiple different http failures

• backout: https://hg.mozilla.org/integration/autoland/rev/6ddefe7e7f25c644dfffc0235e658eb491e86d5d
• push: https://treeherder.mozilla.org/jobs?repo=autoland&group_state=expanded&selectedTaskRun=K31D0PoFTZSxE2MwvZ5L6A.0&revision=4e21d241af64f5d1f3b4796d48ab8d58a961f081
• failure logs:
  • TEST-UNEXPECTED-FAIL | dom/xhr/tests/test_XHR.html | wrong statusText for file_XHR_pass1.xml - got "", expected "OK"
  • TEST-UNEXPECTED-PASS | /wasm/webapi/status.any.worker.html | Response with status 0: compileStreaming - expected FAIL
  • TEST-UNEXPECTED-PASS | /fetch/h1-parsing/lone-cr.window.html | Parsing response with a lone CR before message-body (HTTP/1.1
  • TEST-UNEXPECTED-FAIL | devtools/client/netmonitor/test/browser_net_http3_request_details.js | The status summary value is incorrect. - Got "", expected "OK"

Comment 5

[Thomas Wisniewski [:twisniewski]]

The failures above are trivial to address, with the fallout being from no longer giving http2/http3 status text a default value. However I'm going to wait for feedback from :jdescottes on the devtools test failure, since for that code I don't know whether we want to provide any default text, or go with what the server literally provides.

Comment 6

[Thomas Wisniewski [:twisniewski]]

:jdescottes says it should be fine to just update the tests, as it would be better to be precise about the http2/3 response than add in a default text value, so I'll try landing again today.

Comment 7

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/01ca7125a61d
Enable stricter HTTP response status line parsing; r=valentin,necko-reviewers

Comment 8

[Narcis Beleuzu [:NarcisB]]

Backed out for mochitest failure on test_fetch_basic_http.html

Backout link: https://hg.mozilla.org/integration/autoland/rev/d6e3a985fda44edfa84fae3ecc4b52e6b732f09e
Log link: https://treeherder.mozilla.org/logviewer?job_id=426374350&repo=autoland&lineNumber=6371

Comment 9

[Thomas Wisniewski [:twisniewski]]

Ah, it's a shame that I missed that test in my last push. I'll update the patch and try landing one more time today.

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/e5b2c0e8283f
Enable stricter HTTP response status line parsing; r=valentin,necko-reviewers

Comment 11

[Serban Stanca [:SerbanS]]

https://hg.mozilla.org/mozilla-central/rev/e5b2c0e8283f