HTTP 408 request timeout responses are invisibly repeated |
||||||
Issue descriptionUserAgent: Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/50.0.2661.94 Safari/537.36 Steps to reproduce the problem: 1. Start Fiddler (or a real server) and make any request get a 408 Request Timeout response. 2. fetch() some URL. What is the expected behavior? Multiple Network panel entries, for every retry. What went wrong? A single Network panel entry. Did this work before? N/A Chrome version: 50.0.2661.94 Channel: n/a OS Version: 6.1 (Windows 7, Windows Server 2008 R2) Flash Version: Shockwave Flash 21.0 r0 Fiddler shows multiple requests and responses. chrome:net-internals shows multiple requests but only one response (perhaps problematic as well).
,
May 25 2016
,
May 25 2016
,
May 25 2016
I can confirm this is happening. This is due to the way chromium is architectured. This is caused because Blink dispatches the request to Browser which then dispatches the request to Network Stack. NetStack then sends the request out and then gets back the 408 and decides to retry the request without letting Browser or Blink know about it's retry attempts. Devtools has our hooks in Blink and Browser but not NetStack. Fixing this would require a lot of effort. I am not sure if or when this will be fixed, but I will leave it open. Thanks!
,
Jun 1 2016
I'm seeing a similar issue with GET requests that return a 204. I see the initial request in the Network tab of Devtools (and in our server logs), but if start filtering the Network I start receiving duplicate requests. Could this issue be related?
,
Jun 1 2016
#5 Possible, but more probability of your issue being caused by: http://crbug.com/613188
,
Jun 2 2016
It doesn't look like I have permission to access http://crbug.com/613188.
,
Jun 2 2016
#7 Sorry, I forgot to see if it was restricted. But I believe this issue is caused by the same thing in the linked issue. Sorry, but this issue was posted by an @google and linked some corp stuff so I cannot remove the flag to make it public. I'll post here if/when there's an update.
,
Jul 23 2016
Issue 613188 has been merged into this issue.
,
Sep 21 2017
It looks like this is not something we can solve right now without a huge overhaul to many areas. The netstack is the area this is being resent. The request comes from one process (renderer) then it forwards to browser process. The netstack eventually gets it (and according to spec) a request gets recreated. We could send it back to the renderer to get an "ok", but that would only be useful to devtools and does not seem worth the effort. Thanks for the report, but the good news is that it is surfaced in chrome://net-internals/ |
||||||
►
Sign in to add a comment |
||||||
Comment 1 by phistuck@chromium.org
, May 25 2016