Memory Leak in IE11 even after refreshing the whole page #541

ankurp · 2016-05-31T14:20:41Z

I have created a repo to demonstrate this issue. Just refreshing the page with a basic component keeps increasing memory in IE 11 and it never gets released
https://github.com/ankurp/polymer-webcomponent-IE-memory-leak

robbecker-wf · 2016-05-31T18:24:00Z

I verified this. Seems like to release the memory you need to close all of the IE windows.
EDIT: @ankurp can you try taking out the window.location reload every second and manually refresh? Or set the reload to longer, like 4 seconds?

I tried refreshing manually and the memory leak either isn't there, or doesn't seem as severe. The GC kicks in and handles things.

ankurp · 2016-05-31T21:18:16Z

@robbecker-wf I tried manual refresh and even reduced the frequency of refreshes and memory usage still increases over time. I have seen the garbage collector not kick in after a lot of refreshes.

ebidel · 2016-06-01T15:56:03Z

Can you verify/isolate even more for us? Is it the inclusion of the polyfill alone? Is it adding the Polymer import. Is it creating a DOM module (full or empty)? There's a few things on the page that you can probably get rid of to better reduce the test case.

robbecker-wf · 2016-06-01T16:01:59Z

@ebidel It did not seem to be a problem with JUST the webcomponents polyfill. It needed both to happen. Is that your experience too @ankurp ?

ankurp · 2016-06-01T16:29:09Z

Both are needed as you cannot just use webcomponent with either polymer or webcomponent polyfill in IE. Also this is the simplest example I created. It has nothing but a dom-module https://github.com/ankurp/polymer-webcomponent-IE-memory-leak/blob/master/index.html#L27 and nothing else on the page that can account for 2-3MB increases on every page refresh.

ankurp · 2016-06-01T20:27:46Z

@ebidel I removed everything from the page except for a <h1> tag and including both polymer and webcomponents polyfill script and refreshing causes a memory leak of 2MB https://github.com/ankurp/polymer-webcomponent-IE-memory-leak/blob/master/index.html

ankurp · 2016-06-01T20:30:25Z

@ebidel Also one thing to note is that memory leaks when both scripts are present. If you only include polymer.html or webcomponents polyfill then there is no memory leak but including both will cause a memory leak.

dombre · 2016-06-08T14:42:13Z

Any update on this?

I have the same memory leak as @ankurp.

The most minimal example I'm using is this:

<!DOCTYPE html>
<html>
  <head>

    <script src="https://polygit.org/polymer+:master/components/webcomponentsjs/webcomponents-lite.js"></script>
    <link href="https://polygit.org/polymer+:master/components/polymer/polymer-micro.html" rel="import">

  </head>

  <body>
  </body>
</html>

My conclusions are so far:

It needs both: Webcomponents and Polymer (Micro)
The refresh rate doesn't matter
You see the GC working but it will not free everything. Eventually IE will crash.

ankurp · 2016-06-08T15:31:40Z

@dombre I agree and noticed the same thing. Just including polymer link and webcomponent-lite both together in an empty HTML page causes 2-3MB of memory leak that does not get cleaned up even after several refreshes. Memory is only released when browser or tab is closed.

dombre · 2016-06-08T16:11:01Z

In real world applications this gets worse. Of course, it doesn't matter for SPAs but for elements you include in a 'normal' Website this is bad. Here are some examples:

Example 1: https://www.polymer-project.org/1.0/ will loose ~40 MB every refresh:

Example 2: https://shop.polymer-project.org will loose ~12 MB every refresh:

My Test environments are:

Windows 8.1, IE Version 11.0.9600.18321 (Update Version: 11.0.31)
Windows 10, IE Version 11.306.10586.0 (Update Version: 11.0.31)

For reference, this is the memory consumption of the minimal example #541 (comment) with a 5 secondes refresh rate

ankurp · 2016-06-10T19:31:20Z

Any update on this?

ebidel · 2016-06-14T06:01:43Z

@justinfagnani @azakus mentioned they had looked at memory leaks in IE at one point.

justinfagnani · 2016-06-24T18:09:56Z

Update: @garlicnation is taking a look at this

joao-arvins · 2016-07-11T16:26:39Z

Any updates on this issue?

Also is there any quick patch we can use in the mean time as a temporary fix?
I've tried to use older versions of the library but the problem is still there.

garlicnation · 2016-07-11T18:44:14Z

After extensive experimentation, I've determined that this is definitely an IE issue, not a webcomponents issue. It's possible to get IE11 and Edge to leak memory by creating a variety of different platform-provided objects.

Reported issue: https://developer.microsoft.com/en-us/microsoft-edge/platform/issues/8020622/
The reproduction and bug report in IE Edge as well as 11, so there's a good chance they'll fix it.

I've experimented with a couple of mitigation strategies that show some promise in reducing, if not fixing, this issue. My next step introduce the mitigations into the polymer-element-catalog and determine the effect they have in a real-world application. If one of the patches I come up with has a significant effect, I'll update this bug with it and/or consider patching it into the polyfill until the underlying IE issue is resolved.

kstenson · 2016-07-12T14:00:10Z

@garlicnation can you give more detail of your approach? We have an application using this via x-tags and even refreshing the page is not releasing memory.

I would be happy to test out any patches you have in our app for testing a real life app.

jbotelho2-bb · 2016-07-27T20:29:39Z

@garlicnation any chance you could share what platform-provided objects cause the memory leaks?

We are currently seeing a similar leak on our site, and are in contact with a support engineer at Microsoft. We're having a hard time clarifying that the issue is a bug in IE though, as opposed to an issue in our code or the webcomponents.js code.

felvhage · 2016-08-04T08:49:39Z

Would also be interested in details about mitigating this problem now and any updates on a polyfill solution.

chiefcll · 2016-08-13T14:04:25Z

@garlicnation - Can you provide an update on this issue and if you we're able to come up with any work arounds?

jbotelho2-bb · 2016-08-13T16:48:56Z

For anyone investigating this, it seems to be related to circular references of some sort, but I'm waiting for more info from Microsoft about this. @garlicnation, I think that we will need a workaround/mitigation available, since even if Microsoft comes up with a patch, it won't get pushed out and applied to end-users very quickly.

I also put a workaround in Polymer/polymer#3430 for cases involving a framework like Polymer, where you don't explicitly need HTML imports.

kstenson · 2016-09-13T14:49:33Z

Any update on this issue, we're having real issues with our call centre applications where a new call triggers a new session using a browser reload with querystring params.

Like I said in a previous comment to @garlicnation even a quick post to give use a pointer to where you see this issue arising would be a big help.

ergo · 2016-09-15T18:34:17Z

Any news on this?

bbsteventill · 2016-10-26T14:14:29Z

I have a fix for this; has to do with the document.implementation. From what I found if you have any non-standard attributes/properties that you are putting on a document, IE will leak. I also have submitted a PR for memory leak fixes with in Polymer (Polymer/polymer#3964).

Other things of note, some polymer elements such as paper-dialog will use elements like
iron-overlay-manager; in that code if you look at

Polymer.Gestures.add(document, 'tap', this._onCaptureClick.bind(this));
document.addEventListener('focus', this._onCaptureFocus.bind(this), true);
document.addEventListener('keydown', this._onCaptureKeyDown.bind(this), true);

this may cause a memory leak as well, so you may want to add an event wrapper on document AND window that will remove the events on page unload (the problem is that when you bind, it creates a new fn reference that IE doesnt know how to clean up - I have attached a very aggressive unload script to this issue(Polymer/polymer#3430) )

Anyway like I said I have a PR ready to go that will fix the memory leak with the imports.

I am just waiting for approval to submit a PR. Should be submitted by the end of the week

ergo · 2016-12-13T13:36:18Z

Should this get closed? It seems that #644 fixes this - can we have a confirmation?

kstenson · 2016-12-13T14:44:50Z

@ergo I did a quick check with the latest build from master and was still seeing a memory leak issue. My project doesn't use html imports so i'm not sure the changes in #644 are relevant.

@garlicnation in your initial comments you mentioned a couple of mitigation strategies, do you mind sharing more information?

bbsteventill · 2016-12-13T16:22:27Z

@kstenson you may want to look at this issue: Polymer/polymer#3430
I added this script; it is not perfect but it works for me.
https://github.com/Polymer/polymer/files/480498/polymer_memleak_cleanup.js.txt

Things to be aware of, I had to manually clear out the CustomElements.registry on page unload; also it looks like IE has a memory leak issues with root shadow dom nodes. (for example adding properties directly or putting observers on the root nodes)

kstenson · 2016-12-14T09:19:21Z

@bbsteventill thanks for the link, we aren't using polymer with this lib so we don't have access to CustomElements.registry

bbsteventill · 2016-12-14T19:07:44Z

@kstenson CustomElements is the polyfill in webcomponentsjs that adds support for CustomElements; hense why I suspected/hoped that script or parts of it could help you out.

https://github.com/webcomponents/webcomponentsjs/blob/master/src/CustomElements/register.js

kstenson · 2016-12-15T08:57:15Z

@bbsteventill ah I thought this was a polymer object built on top, thanks for the pointer I will check it out.

kevinpschaaf · 2017-01-26T19:59:11Z

Fix in #644 was reverted in #665 due to a problem with the implementation leading to custom elements in shadow roots being missed by the custom elements polyfill observers. This will need an alternate fix.

liraniom · 2017-03-14T08:28:58Z

Hello, Any update on this?

jdulude · 2017-05-03T12:47:09Z

+1 for an update on this...

Jeff17Robbins · 2017-05-08T00:36:01Z

+1 for an update, please...

xingchch · 2017-06-29T07:35:49Z

+1 for an update, please...

jbotelho2-bb · 2017-06-29T11:30:17Z

We've been using a fork of webcomponentsjs that includes #664 to solve the memory leaks we've been seeing when using Polymer. Since we don't use shadow DOM, this has been fine for us.

Also, at this point it's probably not worth patching version 0.7, since 1.0 is out and will be what most of the webcomponentsjs and Polymer devs will be focusing on. I would hope the webcomponentjs devs would look into this on 1.0 (if it's still an issue), since the world is stuck with IE11 as a supported browser until 2020.

Of course if anyone has their own patch they'd like to submit, by all means open a PR.

jhuesos · 2017-10-30T12:47:01Z

Is this issue solved in WebComponentsJS v1.x?

ttcremers · 2017-10-31T13:38:40Z

Is this issue solved in WebComponentsJS v1.x?

@jhuesos I tested this yesterday and the issue is still there. I was able te reproduce by creating a html document which script includes webcomponentsjs, on every reload it leaks even with an empty body. I tested with webcomponentsjs 1.0.17.

jbotelho2-bb · 2017-10-31T14:02:24Z

May also want to track Polymer/polymer#4511

TimvdLippe · 2018-09-17T17:02:25Z

Thanks for the issue. It has been a great while since this issue was opened. Sadly, this issue misses a live reproduction with the most recent version of the webcomponents polyfill for us to be able to debug and potentially fix. Versions 1 of the polyfills fix a large amount of issues, probably including this one as well.

Please see the CONTRIBUTING.md for guidelines on filing issues. You can start from this jsBin template as a starting point and link it in the issue template under "Live Demo".

Feel free to file a new issue following these instructions if you can still reproduce this issue with versions 1 of the polyfills. Thank you!

AkhilNaidu09 · 2018-11-19T07:43:26Z

Can we implement Garbage collector externally for IE11 using any libraries? I am not sure about this methodology !!! This memory leak issue is not with specific libraries like (Angular, react or any JS), since it is the issue with IE browser, can we use and pure js library to do GC externally?

ankurp mentioned this issue Jun 10, 2016

Memory Leak in IE11 #542

Closed

justinfagnani assigned garlicnation Jun 28, 2016

justinfagnani added bug labels Jun 28, 2016

felvhage mentioned this issue Aug 4, 2016

Polymer will lead to memory leaks in Internet Explorer 11. Polymer/polymer#3430

Closed

kstenson mentioned this issue Oct 26, 2016

Memory Leak in IE11 even after refreshing the whole page x-tag/core#165

Open

notwaldorf added the v0 label Jan 23, 2017

alvarezguille mentioned this issue Apr 7, 2017

Polymer 2 Memory Leak in IE11 Polymer/polymer#4511

Closed

6 tasks

maraist mentioned this issue Oct 25, 2017

IE 11 Memory Leaking on Browser Reload angular/angular#17637

Closed

jhuesos unassigned garlicnation Oct 30, 2017

TimvdLippe closed this as completed Sep 17, 2018

Memory Leak in IE11 even after refreshing the whole page #541

Memory Leak in IE11 even after refreshing the whole page #541

Comments

ankurp commented May 31, 2016

robbecker-wf commented May 31, 2016 • edited Loading

ankurp commented May 31, 2016 • edited Loading

ebidel commented Jun 1, 2016

robbecker-wf commented Jun 1, 2016

ankurp commented Jun 1, 2016

ankurp commented Jun 1, 2016 • edited Loading

ankurp commented Jun 1, 2016

dombre commented Jun 8, 2016

ankurp commented Jun 8, 2016

dombre commented Jun 8, 2016

ankurp commented Jun 10, 2016

ebidel commented Jun 14, 2016

justinfagnani commented Jun 24, 2016

joao-arvins commented Jul 11, 2016

garlicnation commented Jul 11, 2016

kstenson commented Jul 12, 2016 • edited Loading

jbotelho2-bb commented Jul 27, 2016

felvhage commented Aug 4, 2016

chiefcll commented Aug 13, 2016

jbotelho2-bb commented Aug 13, 2016

kstenson commented Sep 13, 2016

ergo commented Sep 15, 2016

bbsteventill commented Oct 26, 2016

ergo commented Dec 13, 2016 • edited Loading

kstenson commented Dec 13, 2016

bbsteventill commented Dec 13, 2016

kstenson commented Dec 14, 2016

bbsteventill commented Dec 14, 2016 • edited Loading

kstenson commented Dec 15, 2016

kevinpschaaf commented Jan 26, 2017

liraniom commented Mar 14, 2017

jdulude commented May 3, 2017

Jeff17Robbins commented May 8, 2017

xingchch commented Jun 29, 2017

jbotelho2-bb commented Jun 29, 2017

jhuesos commented Oct 30, 2017

ttcremers commented Oct 31, 2017

jbotelho2-bb commented Oct 31, 2017

TimvdLippe commented Sep 17, 2018

AkhilNaidu09 commented Nov 19, 2018 • edited Loading

robbecker-wf commented May 31, 2016 •

edited

Loading

ankurp commented May 31, 2016 •

edited

Loading

ankurp commented Jun 1, 2016 •

edited

Loading

kstenson commented Jul 12, 2016 •

edited

Loading

ergo commented Dec 13, 2016 •

edited

Loading

bbsteventill commented Dec 14, 2016 •

edited

Loading

AkhilNaidu09 commented Nov 19, 2018 •

edited

Loading