1077806 - tweetdeck.twitter.com and twitter.com history doesn't load in Nightly 35.0a1 and Aurora 34.0a2 when http2 enabled

Status: RESOLVED WORKSFORME

(Web Compatibility :: Site Reports, defect)

Description

[WildcatRay]

On twitter.com, older timeline items fail to load when scrolling down. Also, clicking buttons/links at top of twitter.com page fail to move page to those pages. Finally, no timeline items load in the columns of tweetdeck.twitter.com.

Site works as expected in Fx 32.0.3.

Comment 1

[Loic]

STR:
Log in https://twitter.com/login or https://tweetdeck.twitter.com/

Result: history doesn't load by scrolling.

Regression range:
http://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=7f81be7db528&tochange=6cbdd4d523a7
Suspected:
Nicholas Hurley — Bug 1047594 - Enable http2 and alpn by default r=mcmanus r=dkeeler 

network.http.spdy.enabled.http2draft=false fixes the issue, similar to bug 1050063.

Comment 2

[Achwaq Khalid]

Confirmed on the latest Nightly (35.0a1 10/03/2014)

Comment 3

[Dirkjan Ochtman (:djc)]

I also get like a 404 Not Found page (simple black text on white bg) with individual tweets like https://twitter.com/adambaldwin/status/518293964571418624 and sometimes twitter.com/ fails to load in a weird way where the tab just sits there and acts like nothing is happening. The devtools shows a sensible request, but with type plain, size 0 and time 0, and the response pane is empty.

Comment 4

[Patrick McManus [:mcmanus]]

wfm.

can someone affected get an http log? That's going to be the interesting detailed info.

https://developer.mozilla.org/en-US/docs/Mozilla/Debugging/HTTP_logging

Comment 5

[Alex]

Attachment: HTTP log Twitter

I had to reload 3 times until the "not found" error appeared. After that, I reloaded another 3 times with the error.

Comment 6

[Loic]

Attachment: log.zip (twitter and tweetdeck)

Comment 7

[Clochix]

I can confirm I have the same behavior: 404 errors that prevent me from doing any action on twitter.com when I'm connected. When I'm not connected, I get random 404 when trying to view status.
It seems that the error disappear when setting network.http.spdy.enabled.http2draft to false.
Setting network.http.spdy.enabled.http2 to true seems to work.

Comment 8

[Bill Gallagher]

This is likely a bug in Twitter's HTTP2 implementation.  We recently enabled a new HTTP2 implementation and it looks like it has an HPACK bug.  We will either fix the bug or temporarily disable HTTP2 early this week.  Sorry for the inconvenience.

Comment 9

[Dave Garrett]

(In reply to Bill Gallagher from comment #8)
> This is likely a bug in Twitter's HTTP2 implementation.

Thanks for the update. I'll switch this to TE, then.

Comment 10

[mozillien]

Same problem with Aurora 34.0a2 too.

Comment 11

[Hallvord R. M. Steen [:hallvors]]

Thanks a lot for commenting, Bill! Hope there's an easy fix waiting for you :)

Comment 12

[Patrick McManus [:mcmanus]]

similar bug filed against chrome - so a server side issue makes sense

https://code.google.com/p/chromium/issues/detail?id=420299

Comment 13

[Achwaq Khalid]

Workaround: toggling "network.http.spdy.enabled.http2draft" off fix the issue

Comment 14

[David Bruant]

Another workaround is to Ctrl+F5 on the urls resulting to "Not found"

Comment 15

[br1.an]

My issue is a little different. https://tweetdeck.twitter.com partially loads, i.e. all my column headers but no content (tweets) therein. I've left it running for 5 mins but doesn't result in a 'page not found' 404 error. Also, there doesn't appear to be any issues on https://twitter.com I just get a completely blank page. Using the workaround, toggling "network.http.spdy.enabled.http2draft" off does resolve the issue.

Furthermore I don't have this issue using Chrome,Version 39.0.2166.2 dev-m (64-bit)

Comment 16

[mozillien]

(In reply to Achwaq Khalid from comment #13)
> Workaround: toggling "network.http.spdy.enabled.http2draft" off fix the issue

Works for me, thanks.

Comment 18

[Achwaq Khalid]

(In reply to mozillien from comment #16)
> (In reply to Achwaq Khalid from comment #13)
> > Workaround: toggling "network.http.spdy.enabled.http2draft" off fix the issue
> 
> Works for me, thanks.

Great :)

Comment 19

[Frederik Braun [:freddy]]

This also affects some b2g branches and pre-releases.

Comment 20

[:gerard-majax]

While I was affected on my Aurora Desktop, I did not noticed any issue on current B2G master.

Comment 21

[Chris Lord [:cwiiis]]

(In reply to Alexandre LISSY :gerard-majax from comment #20)
> While I was affected on my Aurora Desktop, I did not noticed any issue on
> current B2G master.

On the contrary, I was affected on b2g - The official twitter app (their mobile site) refused to load anything except the main feed, and lots of interactions fail, and the third-party twitter app Macaw wouldn't load anything at all. Toggling the pref fixed both for me.

Comment 22

[Karl Dubost💡 :karlcow]

Jan,

could you help us to find the right person at Twitter in charge of HTTP2 implementation?
See comment #8 for the issue.

Comment 23

[Mike Conley (:mconley) (:⚙️)]

I'm on Firefox Desktop running Nightly (35.0a1 (2014-10-05)), and was experiencing this problem. It seems to have cleared up sometime this afternoon. It's possible Twitter has fixed this on their end.

Comment 24

[Patrick McManus [:mcmanus]]

(In reply to Mike Conley (:mconley) - Needinfo me! from comment #23)
> I'm on Firefox Desktop running Nightly (35.0a1 (2014-10-05)), and was
> experiencing this problem. It seems to have cleared up sometime this
> afternoon. It's possible Twitter has fixed this on their end.

from where I sit twitter has disabled their http2 stack this morning and they are speaking spdy/3.1. I presume they will re-enable when the issue is resolved.

Comment 25

[Patrick McManus [:mcmanus]]

issue resolved on server side.. i'll mark it WFM - feel free to resolve it differently if someone feels there is a better match.

Comment 26

[Bill Gallagher]

Just a heads up that we just re-enabled HTTP2 in the datacenter that caused this problem.  I believe that all of the issues are fixed but please let me know if you see any flaky behavior.  Thanks.

Comment 27

[Jan Killian]

(In reply to Bill Gallagher from comment #26)
> Just a heads up that we just re-enabled HTTP2 in the datacenter that caused
> this problem.  I believe that all of the issues are fixed but please let me
> know if you see any flaky behavior.  Thanks.

Just reproduced the issue again with Chrome 40.0.2209.0 dev-m (64-bit). It happened a lot in last months. Can you please test with Chrome-dev too?