Closed
Bug 1114347
Opened 9 years ago
Closed 9 years ago
When loading onedrive.live.com, blank page is returned
Categories
(Core :: JavaScript Engine, defect)
Tracking
()
RESOLVED
FIXED
mozilla37
People
(Reporter: josh.tumath+bugzilla, Assigned: jandem)
References
()
Details
Attachments
(2 files)
231 bytes,
text/html
|
Details | |
10.36 KB,
patch
|
bhackett1024
:
review+
Sylvestre
:
approval-mozilla-aurora+
|
Details | Diff | Splinter Review |
User Agent: Mozilla/5.0 (Windows NT 6.4; Win64; x64; rv:37.0) Gecko/20100101 Firefox/37.0 Build ID: 20141221030204 Steps to reproduce: 1. Signed into my Microsoft account. 2. Went to onedrive.live.com Actual results: A blank page was returned. View source shows that there is page content being returned, but it's not being displayed. Happens regardless of whether e10s is enabled.
Comment 1•9 years ago
|
||
Does this work in Firefox's safe mode? Does this work on release Firefox? Please provide the graphics info from about:support (help > troubleshooting info). Windows 10 is a pre-release OS, do you have a 8.1 or earlier install on the same system? https://support.mozilla.org/en-US/kb/troubleshoot-and-diagnose-firefox-problems#w_3-restart-firefox-in-safe-mode
Flags: needinfo?(josh)
Reporter | ||
Comment 2•9 years ago
|
||
(In reply to Kevin Brosnan [:kbrosnan] from comment #1) > Does this work in Firefox's safe mode? Does this work on release Firefox? > Please provide the graphics info from about:support (help > troubleshooting > info). Windows 10 is a pre-release OS, do you have a 8.1 or earlier install > on the same system? > > https://support.mozilla.org/en-US/kb/troubleshoot-and-diagnose-firefox- > problems#w_3-restart-firefox-in-safe-mode Still occurs in safe mode. Unable to test on release Firefox at the moment but will when I can. I have very slow internet where I am at the moment so it's not worth downloading it. However, I did try on Windows 8.1 in Nightly, and the bug did not occur, which suggests it's a bug in Windows 10. Graphics info: Adapter Description Intel(R) HD Graphics 4000 Adapter Drivers igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32 Adapter RAM Unknown Device ID 0x0166 Direct2D Enabled true DirectWrite Enabled true (6.4.9879.0) Driver Date 10-1-2013 Driver Version 10.18.10.3316 GPU #2 Active false GPU Accelerated Windows 1/1 Direct3D 11 (OMTC) Subsys ID 05891028 Vendor ID 0x8086 WebGL Renderer Google Inc. -- ANGLE (Intel(R) HD Graphics 4000 Direct3D11 vs_5_0 ps_5_0) windowLayerManagerRemote true AzureCanvasBackend direct2d 1.1 AzureContentBackend direct2d 1.1 AzureFallbackCanvasBackend cairo AzureSkiaAccelerated 0
Flags: needinfo?(josh)
Comment 3•9 years ago
|
||
This may be a driver/Windows 10 bug. Reporting it to MS using the Windows 10 feedback app may be useful. http://windowsitpro.com/windows-10/how-supply-feedback-windows-10-technical-preview (hardware and devices area and graphics if that is an option. There is an Intel form at https://servicerequestmgmt.intel.com/webticketui/emailpage.aspx
Comment 4•9 years ago
|
||
This doesn't seem to be Windows 10/driver bug as I also having the same problem on Windows 7 x64. Tested with Safe Mode but still not working... I think it's more likely a website issue as IE11 and Chrome Canary 41 are working properly with OneDrive website. The console also picking up "no elements found" problem... Will try Release version soon.
Comment 5•9 years ago
|
||
Um @Josh Tumath, can you test disable javascript.options.baselinejit in about:config? That seems to fix my problem... @Developers, I leave it to you figure out why the heck that cause the website not to load and stop at blank screen.
Reporter | ||
Comment 6•9 years ago
|
||
(In reply to Jia Yuan Lo from comment #5) > Um @Josh Tumath, > can you test disable javascript.options.baselinejit in about:config? > > That seems to fix my problem... That didn't fix it for me.
Comment 7•9 years ago
|
||
Having the same problem with OneDrive. Disabling javascript.options.baselinejit fixes it for me. Seems to me this is just a workaround as disabling this option surely affects other things. Mozilla/5.0 (Windows NT 6.3; Win64; x64; rv:37.0) Gecko/20100101 Firefox/37.0
Status: UNCONFIRMED → NEW
Ever confirmed: true
Comment 8•9 years ago
|
||
Just discovered that disabling javascript.options.baselinejit adversely affects SunSpider and probably other js benchmarks as well. Went from 100ms to 2000ms. Disabling this pref is not a viable option. Also, note I am on Windows 8.1 so it's not a Windows 10 only bug.
Comment 9•9 years ago
|
||
Disabling baselinejit seems to fix this. Jan have there been any recent baseline changes in 37? To any of the people involved in this bug does this reproduce in Developer edition, beta or release? https://www.mozilla.org/en-US/firefox/developer/ , https://beta.mozilla.org , https://www.mozilla.org/en-US/firefox/
status-firefox34:
--- → ?
status-firefox35:
--- → ?
status-firefox36:
--- → ?
tracking-firefox37:
--- → ?
Component: General → JavaScript Engine
Flags: needinfo?(jdemooij)
Assignee | ||
Comment 10•9 years ago
|
||
Unfortunately I can't reproduce this. I tried with a clean profile and latest Nightly and https://onedrive.live.com/ works fine for me... Does it work with: javascript.options.baselinejit = true javascript.options.ion = false Regression range or STR would be very useful..
Reporter | ||
Comment 11•9 years ago
|
||
(In reply to Jan de Mooij [:jandem] from comment #10) > Unfortunately I can't reproduce this. I tried with a clean profile and > latest Nightly and https://onedrive.live.com/ works fine for me... I was just about to say that it's working for me now, as well, because I just loaded the page and it worked. But then I just tried it again and got a white page again. If you try loading the page in a new tab multiple times, can you reproduce it? (In reply to Jan de Mooij [:jandem] from comment #10) > Does it work with: > > javascript.options.baselinejit = true > javascript.options.ion = false > > Regression range or STR would be very useful.. Nope, it still occurs with ion disabled.
Comment 12•9 years ago
|
||
With a clean profile I am able to get into onedrive.com. When I signin to my onedrive account I just get a blank screen. Eventually I can't even get into onedrive.com. Here's some 'stuff' from my browser console: no element found log:1:1 no element found ScenarioQos.mvc:1:1 lpmanuallogins.length is 0 lpmanuallogins.length is 0 no element found _da.gif:1:1 too much recursion onedrive.live.com:28:760 lpmanuallogins.length is 0 too much recursion onedrive.live.com:28:98 Error: Permission denied to access property '__qosId' skydrivelanding1-5b1d6fb3.js:1:8283 no element found log:1:1 lpmanuallogins.length is 0 Use of getPreventDefault() is deprecated. Use defaultPrevented instead. jquery-1.7.2-d2c5c518.js:17:0 lpmanuallogins.length is 0 no element found _da.gif:1:1 no element found log:1:1 no element found log:1:1 no element found log:1:1 no element found log:1:1 no element found log:1:1 no element found log:1:1
Comment 13•9 years ago
|
||
Jan, the problem is not https://onedrive.live.com/ itself but the problem happens after you signed in... You can login through here: https://onedrive.live.com/?gologin=1&mkt=en-US Also, the problem seems to be cross-platform as I did try checking the site with Nightly Android with Request Desktop Page enabled and still blank page after signed in. So far, release version 34 and beta 35 are not affected. Will try Developer Edition soon.
Assignee | ||
Comment 14•9 years ago
|
||
(In reply to Josh Tumath from comment #11) > If you try loading the page in a new tab multiple times, can you reproduce > it? Nope :( (In reply to Jia Yuan Lo from comment #13) > Jan, the problem is not https://onedrive.live.com/ itself but the problem > happens after you signed in... > > You can login through here: https://onedrive.live.com/?gologin=1&mkt=en-US Yes, I did sign in but it loads fine. Do you have many files there? The account I use is my test account, maybe I should upload a few things...
Assignee | ||
Comment 15•9 years ago
|
||
(In reply to Gary [:streetwolf] from comment #12) > too much recursion onedrive.live.com:28:98 This one is suspicious because recursion limits are platform dependent. I tried on OS X and our recursion limit there is much higher than on other platforms. I'll try this on Windows later.
Comment 16•9 years ago
|
||
I could reproduce using: 1) Create a new profile 2) Go to: https://onedrive.live.com/?gologin=1&mkt=en-US 3) Login with hotmail account 4) After logging in, wait till your documents are loaded 5) Open new tab (ctrl+t) 6) Visit https://onedrive.live.com/ Expected result: View same page as in (4) Actual result: - Empty page - Too much recursion error in console
Comment 17•9 years ago
|
||
Seems there are different scenarios to this problem using a new profile. In my case going to https://onedrive.live.com/?gologin=1&mkt=en-US produces a blank page. However, going to https://onedrive.live.com/ works but eventually it also produces a blank page after a few tries. It might do this after I enter my login information and get a blank page.
Comment 18•9 years ago
|
||
I could reproduce this on 64bit Firefox on FF34/35/36 and 37
status-firefox37:
--- → affected
Comment 19•9 years ago
|
||
I cannot reproduce the problem on Nightly37.0a1 32bit build. https://hg.mozilla.org/mozilla-central/rev/8850aa0f5332 Mozilla/5.0 (Windows NT 6.1; WOW64; rv:37.0) Gecko/20100101 Firefox/37.0 ID:20141229030217 But, I can reproduce on on Nightly37.0a1 x64 build. https://hg.mozilla.org/mozilla-central/rev/8850aa0f5332 Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:37.0) Gecko/20100101 Firefox/37.0 ID:20141229030217 So, I guess this is 64bit build specific problem.
Hardware: x86 → x86_64
Comment 20•9 years ago
|
||
Steps 1. Open https://onedrive.live.com/ and Click [Sign in] 2. Type e-mail and Click [Next] 3. Type password and Click [Sign in] 4. Open new tab(Ctrl+T) 5. Open https://onedrive.live.com/ 6. Reload (F5) 3-10 times Regression window(m-c 64bit) Good: http://hg.mozilla.org/mozilla-central/rev/4646259ab62d Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:27.0) Gecko/20100101 Firefox/27.0 ID:20131028030205 Bad: http://hg.mozilla.org/mozilla-central/rev/518f5bff0ae4 Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:28.0) Gecko/20100101 Firefox/28.0 ID:20131029030201 Pushlog: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=4646259ab62d&tochange=518f5bff0ae4
Comment 21•9 years ago
|
||
Regarding your regression range Alice, I can swear OneDrive was working for me well after the 'bad' pushlog.
Comment 22•9 years ago
|
||
(In reply to Gary [:streetwolf] from comment #21) > Regarding your regression range Alice, I can swear OneDrive was working for > me well after the 'bad' pushlog. I am not sure, but, I guess the site changed something between 2013-10-29 and your first notice.
Comment 23•9 years ago
|
||
In local non-pgo 64 bit build(using msvc2012), I can also reproduce the problem on the build from the good changeset of comment#20..... :/
Comment 24•9 years ago
|
||
Fwiw, this also happens on 35beta, OpenBSD/i386 and amd64. So not windows-specific, nor 64-bits only.
Comment 25•9 years ago
|
||
And this is not related to login or not, i can reproduce it when loading a link i've been given with an authkey param to see a private album in a live.com account. I dont have a live account anyway...
Assignee | ||
Comment 26•9 years ago
|
||
I can reproduce this on Linux with a 64-bit build. When we throw the overrecursion error, there are about 550 frames on the stack. The max recursion depth is better when we can stay in JIT code and avoid the huge Interpret frames on the stack. Unfortunately, here a particular script is not JITted, due to the UseNewTypeForClone optimization: we create many clones and each has a separate usecount, so we never enter Baseline :( Below is a small testcase that prints 80 on OS X. If I make UseNewTypeForClone return false, we get 3000-5000 (because we can enter Ion, but with --no-ion I get 156 which is also a lot more). If I make UseNewTypeForClone return false on Linux, I can no longer reproduce the problem. Brian, any thoughts on this? It seems like overzealous cloning can also be bad for memory usage, it'd be nice if we could limit it somehow. function f(x) { var clone = function() { f.apply(null, arguments); }; print(x); clone(x + 1); } f(0);
Flags: needinfo?(jdemooij) → needinfo?(bhackett1024)
Assignee | ||
Comment 27•9 years ago
|
||
We also run the arguments analysis for each script clone I think...
Assignee | ||
Comment 28•9 years ago
|
||
On OS X I get 808, on Linux 102. Disabling UseNewTypeForClone improves it to ~80000 (!) on OS X and ~2800 on Linux.
Updated•9 years ago
|
Comment 29•9 years ago
|
||
(In reply to Jan de Mooij [:jandem] from comment #26) > Brian, any thoughts on this? It seems like overzealous cloning can also be > bad for memory usage, it'd be nice if we could limit it somehow. Well, the UseNewTypeForClone is designed to help with constructor factories like the classes in Prototype.js, and if a complex page has hundreds of classes made by the factory then there could legitimately be hundreds of functions we would want to track separately. Now, there isn't any need for this mechanism when the function is being called rather than constructed; is that the case on this page? It should be fine to disable UseNewTypeForClone on a function whenever one of the clones has been called rather than constructed, though implementing that might be tricky as I don't think the clones have any pointers back to the function they were cloned from.
Flags: needinfo?(bhackett1024)
Assignee | ||
Comment 30•9 years ago
|
||
(In reply to Brian Hackett (:bhackett) from comment #29) > Now, there isn't any need for this mechanism when the function is being > called rather than constructed; is that the case on this page? It should be > fine to disable UseNewTypeForClone on a function whenever one of the clones > has been called rather than constructed, though implementing that might be > tricky as I don't think the clones have any pointers back to the function > they were cloned from. Yeah, OK. It might be simpler to check for presence of `this` (likely constructor) or `return` (likely not a constructor). Both would work for this site as far as I can tell. I'll implement the `this` one and see if it works.
Flags: needinfo?(jdemooij)
Assignee | ||
Comment 31•9 years ago
|
||
Fixes the UseNewTypeForClone heuristic to check for the presence of `this` as well. This should fix the website and also wins about 10% on box2d, it has a very similar function.
Assignee: nobody → jdemooij
Status: NEW → ASSIGNED
Flags: needinfo?(jdemooij)
Attachment #8544570 -
Flags: review?(bhackett1024)
Comment 32•9 years ago
|
||
Comment on attachment 8544570 [details] [diff] [review] Patch Review of attachment 8544570 [details] [diff] [review]: ----------------------------------------------------------------- Nice!
Attachment #8544570 -
Flags: review?(bhackett1024) → review+
Assignee | ||
Comment 33•9 years ago
|
||
[Tracking Requested - why for this release]: https://hg.mozilla.org/integration/mozilla-inbound/rev/4a979ea519e2
tracking-firefox36:
--- → ?
Assignee | ||
Comment 34•9 years ago
|
||
Alice (or anybody else who could repro), can you confirm inbound revision 4a979ea519e2 fixes this?
Flags: needinfo?(alice0775)
Comment 35•9 years ago
|
||
I can verify that the problem was fixed. Reproduced the problem: https://hg.mozilla.org/integration/mozilla-inbound/rev/b42615e51c81 Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:37.0) Gecko/20100101 Firefox/37.0 ID:20150106080349 I cannot reproduce the problem any more(I tried five times) https://hg.mozilla.org/integration/mozilla-inbound/rev/4a979ea519e2 Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:37.0) Gecko/20100101 Firefox/37.0 ID:20150106081241
Flags: needinfo?(alice0775)
Assignee | ||
Comment 36•9 years ago
|
||
(In reply to Alice0775 White from comment #35) > I can verify that the problem was fixed. Great, thanks!
Comment 37•9 years ago
|
||
Problem indeed fixed. Thanks.
Comment 38•9 years ago
|
||
Err... How do I test? Where can I get the patch? Do I need to compile myself? :S I'm still new to this Bugzilla world...
https://hg.mozilla.org/mozilla-central/rev/4a979ea519e2
Status: ASSIGNED → RESOLVED
Closed: 9 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla37
Comment 40•9 years ago
|
||
(In reply to Jia Yuan Lo from comment #38) > Err... How do I test? Where can I get the patch? Do I need to compile > myself? :S > I'm still new to this Bugzilla world... (In reply to Jia Yuan Lo from comment #38) > Err... How do I test? Where can I get the patch? Do I need to compile > myself? :S > I'm still new to this Bugzilla world... You can download latest tinderbox m-c build. http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-central-win32/ http://ftp.mozilla.org/pub/mozilla.org/firefox/tinderbox-builds/mozilla-central-win64/
Assignee | ||
Comment 41•9 years ago
|
||
Comment on attachment 8544570 [details] [diff] [review] Patch Approval Request Comment [Feature/regressing bug #]: Older JIT bug. [User impact if declined]: Microsoft OneDrive doesn't work on some platforms (Windows, Linux). [Describe test coverage new/current, TBPL]: Fixes the issue, code is tested on TBPL. [Risks and why]: No risk, patch just tweaks some heuristics. [String/UUID change made/needed]: None.
Attachment #8544570 -
Flags: approval-mozilla-aurora?
Updated•9 years ago
|
Updated•9 years ago
|
Attachment #8544570 -
Flags: approval-mozilla-aurora? → approval-mozilla-aurora+
Comment 43•9 years ago
|
||
Hello, I know this is an old thread, but I have exactly the same issue as is mentioned in this thread. Is there any way this could be possibly fixed? -Xander
Assignee | ||
Comment 44•9 years ago
|
||
(In reply to xanderlenstra from comment #43) > Hello, > I know this is an old thread, but I have exactly the same issue as is > mentioned in this thread. > Is there any way this could be possibly fixed? > -Xander Which OS and Firefox version? Can you reproduce this with a clean profile?
Comment 45•9 years ago
|
||
I'm using Firefox 41.0.2 on Windows 7 64 bits. When I log off of one-drive I can't even open the https://onedrive.live.com/ page, because it's just plain blank. When I try it on a different computer or browser it doesn't seem to happen, although I haven't tested it recently. -Xander
You need to log in
before you can comment on or make changes to this bug.
Description
•