1269170 - KanColle, a popular Flash game doesn't work with Firefox 46, raising a connection error

Status: VERIFIED DUPLICATE of bug 1269055

(Core :: Networking: HTTP, defect)

Description

[Kohei Yoshino [:kohei]]

Japanese Firefox users are reporting that a popular Flash game called 艦これ (KanColle) [1] doesn't work well with Firefox 46, often raising a connection error. They say downgrading to Firefox 45 solved the error.

Firefox 46 has made some Japanese input method changes but I don't think those are related. A possible culprit might be disabling async plugin init in Bug 1241668 but I'm not sure.

Looking for a tester.

[1] https://en.wikipedia.org/wiki/Kantai_Collection

Comment 1

[Kohei Yoshino [:kohei]]

One KanColle user kindly helped test the issue by flipping the dom.ipc.plugins.asyncInit.enabled pref to true on Firefox 46, but he said the error persisted. Any thoughts?

Comment 2

[Xidorn Quan [:xidorn] UTC+11]

It may need an engineer in Japan to debug it, or alternatively use a Japan VPN. As far as I know, KanColle only accepts Japanese IP.

Comment 3

[Kohei Yoshino [:kohei]]

AFAIK Mozilla Japan office is closed this week due to public holidays:
https://en.wikipedia.org/wiki/Golden_Week_%28Japan%29

Comment 4

[(No longer employed by Mozilla) Aaron Klotz]

Async plugin init has not been enabled on release since last August and has been disabled in all prerelease channels since January - it is highly unlikely that this is async init related.

Comment 5

[Benjamin Smedberg]

Can we get a tester to use mozregression to find a nightly regression range?

Comment 6

[Michelle Funches - QA]

I have created an account with http://kancolle.gogames.me/server, I will get familiar and let you know what I find out.
Version 	46.0
Build ID 	20160421124000
User Agent 	Mozilla/5.0 (Windows NT 10.0; WOW64; rv:46.0) Gecko/20100101 Firefox/46.0
If anyone is familiar with this game, please give it a go...

Comment 7

[Xidorn Quan [:xidorn] UTC+11]

I've used some tricks to login... but it seems all its servers are full currently, so new registers are impossible. Probably need someone who is already playing that game to test.

Comment 8

[Kohei Yoshino [:kohei]]

Asked a KanColle player hATrayflood for help.

Comment 9

[ABE Hiroki (:hATrayflood)]

I will play KanColle all day tomorrow, using Firefox 45 with Flash Player 21.0.0.213.

(In reply to Michelle Funches - QA from comment #6)
> I have created an account with http://kancolle.gogames.me/server, I will get
> familiar and let you know what I find out.
maybe http://kancolle.gogames.me/server is a unofficial imitation plagiarism site...

official site is http://www.dmm.com/netgame/feature/kancolle.html
but, you can't start new game, because server is full.

Comment 10

[Michelle Funches - QA]

Acknowledgement & Confirmation: Thank you Abe & Kohei: Yes, this is does limit my ability to test glad there is some one to step in.

Comment 11

[ABE Hiroki (:hATrayflood)]

(In reply to ABE Hiroki (:hATrayflood) from comment #9)
> I will play KanColle all day tomorrow, using Firefox 45 with Flash Player
> 21.0.0.213.
oh, sorry, misstyped. using Firefox 46.

Comment 12

[ABE Hiroki (:hATrayflood)]

network error reproduced 2 times while 2 hours.

OS X		10.11.4
Firefox		46.0.1
Flash Player	21.0.0.213

it seems flash object sent post request, but response is blank.
I'll add some screenshots.

Comment 13

[ABE Hiroki (:hATrayflood)]

Attachment: post_slotset.png

Comment 14

[ABE Hiroki (:hATrayflood)]

Attachment: post_slotset_response.png

Comment 15

[ABE Hiroki (:hATrayflood)]

Attachment: post_slotset_timing.png

Comment 16

[ABE Hiroki (:hATrayflood)]

Attachment: post_questlist.png

Comment 17

[ABE Hiroki (:hATrayflood)]

Attachment: post_questlist_cookie.png

Comment 18

[ABE Hiroki (:hATrayflood)]

Attachment: post_questlist_header.png

Comment 19

[ABE Hiroki (:hATrayflood)]

Attachment: post_questlist_parameter.png

Comment 20

[ABE Hiroki (:hATrayflood)]

Attachment: post_questlist_response.png

Comment 21

[ABE Hiroki (:hATrayflood)]

Attachment: post_questlist_timing.png

Comment 22

[ABE Hiroki (:hATrayflood)]

I got same error while using safari.

OS X		10.11.4
Safari		9.1
Flash Player	21.0.0.213

Comment 23

[ABE Hiroki (:hATrayflood)]

Attachment: safari_post_preset_deck.png

Comment 24

[ABE Hiroki (:hATrayflood)]

Attachment: safari_post_preset_deck_resource.png

Comment 25

[Kohei Yoshino [:kohei]]

How about Chrome and Firefox 46?

Comment 26

[ABE Hiroki (:hATrayflood)]

(In reply to Kohei Yoshino [:kohei] from comment #25)
> How about Chrome and Firefox 46?

Firefox 46:
(In reply to ABE Hiroki (:hATrayflood) from comment #12)
> network error reproduced 2 times while 2 hours.
> 
> OS X		10.11.4
> Firefox		46.0.1
> Flash Player	21.0.0.213
> 
> it seems flash object sent post request, but response is blank.
> I'll add some screenshots.

Chrome 50:
now testing.

Comment 27

[Kohei Yoshino [:kohei]]

Ah, I meant Firefox *45*. People are downgrading from 46 to solve the issue.

Comment 28

[ABE Hiroki (:hATrayflood)]

network error not reproduced using Chrome 50.
now testing Firefox *45*.

Comment 29

[ABE Hiroki (:hATrayflood)]

network error not reproduced using Firefox *45.1.1esr*.

Comment 30

[Liz Henry (:lizzard) (relman/hg->git project)]

Are we concluding this is a regression from 46? 
Maybe the network team should look rather than plugins. Jason can you suggest someone who could look into this issue ?

Comment 31

[Liz Henry (:lizzard) (relman/hg->git project)]

Kohei is there someone we should ask to investigate for bugs marked jp-critical?

Comment 32

[Kohei Yoshino [:kohei]]

hATrayflood: can you please use mozregression to find a regression range? People are still complaining about this on Twitter, so it's not a temporary server-side issue.

Comment 33

[Kohei Yoshino [:kohei]]

(In reply to Liz Henry (:lizzard) (needinfo? me) from comment #31)
> Kohei is there someone we should ask to investigate for bugs marked
> jp-critical?

Mozilla engineers in Japan might help, but I guess no one there is actively playing this Flashy game.

Comment 34

[[e-mail]]

(web compatibility) Michael, KanColle is an extremely popular game in Japan's young generation, could you find people take a look on that?

Comment 35

[Mike Taylor [:miketaylr]]

Karl is based in Japan, so he should be able to help us find a regression range.

Karl -- ping me if you don't know how to use mozregression (docs are very good though, http://mozilla.github.io/mozregression/).

Comment 36

[Jason Duell]

A regression range would be super-helpful here.

It would also be great to have an HTTP log from someone who can play the game (sounds like they have to be in Japan?).  A log may give us an idea of whether this is a necko issue, etc.

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

Sounds like hATrayflood is our best hope for both of these right now.  

Please needinfo me if/when we get a regression range or an http log.

Comment 37

[Karl Dubost💡 :karlcow]

Happy to learn about mozregression, less happy to have install Flash ;)
OK let's do that.

Diving into the http://mozilla.github.io/mozregression/ documentation.
If I'm unsuccessful I will be at the Tokyo Office tomorrow. 

hmmm…
pip install --user -U mozregression
…
Successfully installed BeautifulSoup-3.2.1 PyHawk-with-a-single-extra-commit-0.1.5 blessings-1.6 colorama-0.3.7 configobj-5.0.6 mozdevice-0.47 mozfile-1.2 mozinfo-0.9 mozinstall-1.12 mozlog-3.1 moznetwork-0.27 mozprocess-0.22 mozprofile-0.28 mozregression-2.3.5 mozrunner-6.11 mozversion-1.4 redo-1.5 six-1.10.0 slugid-1.0.7 taskcluster-0.0.32

Retrospectively I should have created a virtualenv before. Rah, then bah… ^_^

OK flash installed.

Game URL
http://www.dmm.com/netgame/social/-/gadgets/=/app_id=854854/

The game is working well in OSX 10.11.4 (15E65)
with Flash 21.0.0.226
and Firefox Nightly 49.0a1 (2016-05-10)

Trying with 
mozregression --good 45 --bad 46

**********
You should use a config file. Please use the --write-config command line flag to help you create one.
**********

 0:05.65 INFO: Using date 2016-01-25 for release 46
 0:05.65 INFO: Using date 2015-12-14 for release 45
 0:12.67 INFO: Testing good and bad builds to ensure that they are really good and bad...
 0:12.68 INFO: Downloading build from: https://archive.mozilla.org/pub/firefox/nightly/2015/12/2015-12-14-12-22-42-mozilla-central/firefox-46.0a1.en-US.mac.dmg
===== Downloaded 100% =====
 0:35.81 INFO: Running mozilla-central build for 2015-12-14
 1:11.46 INFO: Launching /private/var/folders/bg/lmsn8jdx47748dbg0nrh2m8h0000gn/T/tmpF3D620/FirefoxNightly.app/Contents/MacOS/firefox
 1:11.50 INFO: application_buildid: 20151214122242
 1:11.50 INFO: application_changeset: 06ab805e14b5e6d8a8a99c7846c8fde163a9e5d7
 1:11.50 INFO: application_name: Firefox
 1:11.50 INFO: application_repository: https://hg.mozilla.org/mozilla-central
 1:11.50 INFO: application_version: 46.0a1
Was this nightly build good, bad, or broken? (type 'good', 'bad', 'skip', 'retry' or 'exit' and press Enter): good
 6:49.87 INFO: Using local file: /var/folders/bg/lmsn8jdx47748dbg0nrh2m8h0000gn/T/tmpcFsq7c/2016-01-25--mozilla-central--firefox-47.0a1.en-US.mac.dmg (downloaded in background)
 6:49.87 INFO: Running mozilla-central build for 2016-01-25
 7:29.40 INFO: Launching /private/var/folders/bg/lmsn8jdx47748dbg0nrh2m8h0000gn/T/tmpnZnOTJ/FirefoxNightly.app/Contents/MacOS/firefox
 7:29.43 INFO: application_buildid: 20160125060632
 7:29.43 INFO: application_changeset: 3f41d7d0f544ebd98273e39bd945c28878a47427
 7:29.43 INFO: application_name: Firefox
 7:29.43 INFO: application_repository: https://hg.mozilla.org/mozilla-central
 7:29.43 INFO: application_version: 47.0a1
Was this nightly build good, bad, or broken? (type 'good', 'bad', 'skip', 'retry' or 'exit' and press Enter): good
 9:12.09 ERROR: Build was expected to be bad! The initial good/bad range seems incorrect.


So yes indeed no issue on my side. But maybe it doesn't happen always. It would be interesting when/what part of the game are likely to raise the error. 

I could ask tomorrow at the office.

Comment 38

[Mike Taylor [:miketaylr]]

Kohei, can you provide STR for Karl? (if they exist) Thanks.

Comment 39

[Kohei Yoshino [:kohei]]

I have no idea. As :xidorn said earlier, the game site is not accepting new registrations due to a high demand, so I cannot test it myself even if I set up a VPN from Canada. :hATrayflood is an active player who might be able to help here.

Comment 40

[Karl Dubost💡 :karlcow]

hATrayflood 
if you are upgrading the flash version do you still have the issue?

Comment 41

[Mike Taylor [:miketaylr]]

OK, thanks kohei. 

(Karl also has an account now and can help test, but apparently can't reproduce. Some clarity around STR would be good.)

Comment 42

[ABE Hiroki (:hATrayflood)]

(In reply to Karl Dubost :karlcow from comment #40)
> hATrayflood 
> if you are upgrading the flash version do you still have the issue?

now testing using Firefox 46.0.1 with Flash Player 21.0.0.226 on OSX 10.11.4

Comment 43

[ABE Hiroki (:hATrayflood)]

Attachment: bug1269170.log

(In reply to Karl Dubost :karlcow from comment #40)
> hATrayflood 
> if you are upgrading the flash version do you still have the issue?
network error reproduced 5 times while 3 hours.
this issue is stiil active.

(In reply to Jason Duell [:jduell] (needinfo? me) from comment #36)
> It would also be great to have an HTTP log from someone who can play the
> game (sounds like they have to be in Japan?).  A log may give us an idea of
> whether this is a necko issue, etc.
> 
>    https://developer.mozilla.org/en-US/docs/Mozilla/Debugging/HTTP_logging
I add a log.

Comment 44

[Xidorn Quan [:xidorn] UTC+11]

It is probably not a good idea to publish the log here, since that includes sensitive infomation like cookies... Probably this should be sent privately. And you may want to change your password as soon as possible to avoid anyone use the cookies you published here to do something to your game account.

You mentioned that the network error happens 5 times in 3 hours, but the log only contains two minutes. Do you see the network error within this two minutes time? I didn't find either "questlist" or "slotset" in this log.

Comment 45

[ABE Hiroki (:hATrayflood)]

(In reply to Xidorn Quan [:xidorn] (UTC+10) from comment #44)
> It is probably not a good idea to publish the log here, since that includes
> sensitive infomation like cookies... Probably this should be sent privately.
> And you may want to change your password as soon as possible to avoid anyone
> use the cookies you published here to do something to your game account.
> 
oh, sorry. thank you for your advise.
I changed password and re-loged in.

> You mentioned that the network error happens 5 times in 3 hours, but the log
> only contains two minutes. Do you see the network error within this two
> minutes time? I didn't find either "questlist" or "slotset" in this log.
log includes last network error in today.
error request is /kcsapi/api_get_member/preset_deck

Comment 46

[Xidorn Quan [:xidorn] UTC+11]

Thanks for the help. ni? Jason again since there are HTTP logs now.

Comment 47

[Marcus Christie]

This sounds like a problem that I've run into with a Flash application developed using the Flex framework.  It seems to primarily affect POST requests with large request bodies.  The bug is absent from Firefox 45, but present in version 46. I also saw the bug in Firefox Developer Edition 47 but not in Firefox Developer Edition 48.

In Flash, this manifests as "Error #2032".  Is that the same error that this KanColle game is getting?  I guess I'm just not sure if this is the same bug I'm seeing.

Comment 48

[Masatoshi Kimura [:emk]]

Marcus, if you can reproduce the problem reliably, could you find a regression range using mozregression?

Comment 49

[Marcus Christie]

Yeah I can reliably reproduce the problem. I ran mozregression like so

mozregression --good 45 --bad 46

And after several trials it ends up with this

83:54.30 INFO: Running mozilla-inbound build built on 2016-01-12 08:50:29.061000, revision e74d457b
83:56.19 INFO: Launching c:\users\mchris~1\appdata\local\temp\tmpfhfp9n\firefox\firefox.exe
83:56.22 INFO: application_buildid: 20160111235255
83:56.23 INFO: application_changeset: e74d457b8b1100479d6e011b7658ef0baa31c3d2
83:56.23 INFO: application_name: Firefox
83:56.23 INFO: application_repository: https://hg.mozilla.org/integration/mozilla-inbound
83:56.23 INFO: application_version: 46.0a1
Was this inbound build good, bad, or broken? (type 'good', 'bad', 'skip', 'retry', 'back' or 'exit' and press Enter): bad
85:00.97 INFO: Narrowed inbound regression window from [39e69992, 5737b016] (3 revisions) to [39e69992, e74d457b] (2 revisions) (~1 steps left)
85:00.97 INFO: Oh noes, no (more) inbound revisions :(
85:00.97 INFO: Last good revision: 39e699922ca7763e81c690e706a451f0d849059f
85:00.97 INFO: First bad revision: e74d457b8b1100479d6e011b7658ef0baa31c3d2
85:00.97 INFO: Pushlog:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=39e699922ca7763e81c690e706a451f0d849059f&tochange=e74d457b8b1100479d6e011b7658ef0baa31c3d2

I'll attach the full log as well in case it is useful.

Comment 50

[Marcus Christie]

Attachment: log of my mozregression session

Comment 51

[Marcus Christie]

Tomorrow I'll try the --find-fix option as well since the bug isn't there in Firefox Dev Edition (48).

Comment 52

[Xidorn Quan [:xidorn] UTC+11]

Bug 1236277 looks to be the related one. And if that is not reproducible with 48, my guess is that it is probably the same issue as bug 1269055 which has been fixed ~two weeks ago on aurora, and should probably have been fixed in beta as well. You can probably use mozregression to confirm that.

hATrayflood, could you try the latest Firefox Dev Edition 48 and see if this issue still happens there?

Comment 53

[Xidorn Quan [:xidorn] UTC+11]

Adding bug 1269055 to see also, as it is suspected to be the fix of this bug.

Comment 54

[Marcus Christie]

Here's the end of the mozregression output when called with

  mozregression.exe --bad 46 --find-fix

22:54.54 INFO: Narrowed inbound regression window from [2872b905, 23bf1e82] (4 revisions) to [2872b905, aef562ea] (3 revisions) (~1 steps left)
22:54.54 INFO: Pushlog:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=2872b9050cab51f33bc0d95b72461a990b80dd34&tochange=aef562eaad80d8c5bc4b1ba97c653e71593c6b79

22:54.54 INFO: Using local file: c:\users\mchris~1\appdata\local\temp\tmplqf_fy\ff82b00b481e--mozilla-inbound--firefox-49.0a1.en-US.win64.zip (downloaded in background)
22:54.54 INFO: Running mozilla-inbound build built on 2016-05-04 02:42:28.036000, revision ff82b00b
22:57.22 INFO: Launching c:\users\mchris~1\appdata\local\temp\tmprfddic\firefox\firefox.exe
22:57.25 INFO: application_buildid: 20160503174137
22:57.25 INFO: application_changeset: ff82b00b481ed96257bd93ce3e28e882a2f292b6
22:57.25 INFO: application_name: Firefox
22:57.25 INFO: application_repository: https://hg.mozilla.org/integration/mozilla-inbound
22:57.25 INFO: application_version: 49.0a1
Was this inbound build good, bad, or broken? (type 'good', 'bad', 'skip', 'retry', 'back' or 'exit' and press Enter): bad
23:17.75 INFO: Narrowed inbound regression window from [2872b905, aef562ea] (3 revisions) to [ff82b00b, aef562ea] (2 revisions) (~1 steps left)
23:17.75 INFO: Oh noes, no (more) inbound revisions :(
23:17.75 INFO: First good revision: aef562eaad80d8c5bc4b1ba97c653e71593c6b79
23:17.75 INFO: Last bad revision: ff82b00b481ed96257bd93ce3e28e882a2f292b6
23:17.75 INFO: Pushlog:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=ff82b00b481ed96257bd93ce3e28e882a2f292b6&tochange=aef562eaad80d8c5bc4b1ba97c653e71593c6b79

(mozregression)

Comment 55

[Marcus Christie]

I also confirmed that this bug is fixed for me in the latest Firefox 47 beta, 47.0b6.

Comment 56

[Kohei Yoshino [:kohei]]

Thank you very much Marcus! So this is definitely a dup of Bug 1269055.

Comment 57

[Kohei Yoshino [:kohei]]

hATrayflood: can you verify that your issue has been solved with Firefox 47 Beta?

Comment 58

[ABE Hiroki (:hATrayflood)]

now testing using Firefox 47.0b7 with Flash Player 21.0.0.242 on OSX 10.11.5

Comment 59

[ABE Hiroki (:hATrayflood)]

network error not reproduced using Firefox 47.0b7. it seems to be solved this issue.

Comment 60

[Kohei Yoshino [:kohei]]

Thank you!

Comment 61

[Liz Henry (:lizzard) (relman/hg->git project)]

If we have any way to communicate with the server admins for this game site, here's a suggestion from our developers for a workaround:  Adjust your server side persistent connection timeout to be about 5 seconds longer than the firefox value. Firefox uses 115 seconds for HTTP/1.x and 180 seconds for HTTP/2.  

Hope that helps.  I know this isn't ideal for these game users in Japan, but we are leaning against doing a 46.0.2 release as discussed in bug 1269055. The issue is fixed in 47 which is planned to release on June 7th.