Hi Martin,

I've identified what I believe to be the root cause (touchevents on nodes
stamped from a template + edge case timing of the removal of a node that is
a tap target) and reported the crash case to Apple (
https://bugs.webkit.org/show_bug.cgi?id=136506)

The fix in Polymer is going to take me some more time to complete.

In the mean time, I've filed
https://github.com/Polymer/polymer-gestures/issues/56 to track for myself.

Thanks!


On Thu, Sep 4, 2014 at 1:06 AM, Martin Kleinschrodt <
[email protected]> wrote:

> Thanks, guys! I really appreciate the help. Glad, Daniel could reproduce
> this and it's not something stupid I did.
>
> For what it's worth, I think your analysis is pretty spot on. It seems
> there is some race condition going on since the problem occurs only part of
> the time. All things considered, I now think the problem was most likely
> introduced in polymer-gestures since TemplateBinding didn't see a lot of
> changes between 0.3.4 and 0.3.5.
>
> On 3. September 2014 at 23:14:40, 'Daniel Freedman' via Polymer (
> [email protected]) wrote:
>
> I was able to make a repro case just with templates and tap handlers.
>
> It looks like this error is triggered when the a node with a touch handler
> is being removed by a template model update while a touch event is being
> dispatched.
>
> Minimal repro: https://gist.github.com/azakus/72d2b7d1ffde1e94283a
>
> Since this seems related to template stamping timing, I'll have to sit
> down with the template polyfill authors to get a better understanding of
> the exact timing of the polyfill.
>
> Thanks for your patience.
>
>
> On Wed, Sep 3, 2014 at 10:23 AM, Daniel Freedman <[email protected]>
> wrote:
>
>> Hi Martin,
>>
>> I'm not very familiar with iOS, but it looks like there was a bad
>> dereference inside of the WebView (that 0xfffffff9 looks like the highest
>> address on your process stack).
>>
>> I did get it to fail in the ios simulator as well, but mobile safari only
>> reports a "bus error", which suggests that this is a bad pointer
>> dereference.
>>
>> I have a hunch that this is related to a node being removed from the DOM
>> that also has a touch listener on it, but that idea is just based on the
>> fact that safari stops sending touch events when the origin node is removed.
>>
>> The WebCore::Document::touchEventsChangedTimerFired code is an IOS only
>> extension to WebKit, so I'm not sure how to look at what that is doing.
>>
>>
>> On Wed, Sep 3, 2014 at 9:46 AM, Rob Dodson <[email protected]> wrote:
>>
>>> Hey Martin, sorry for the delay, I think we're still catching up a bit
>>> from the long weekend :)
>>> We're looking into this and hopefully will have a response soon.
>>>
>>> - Rob
>>>
>>>
>>> On Wednesday, September 3, 2014 7:29:18 AM UTC-7, Martin Kleinschrodt
>>> wrote:
>>>>
>>>> Anyone? I could really use some help on this...
>>>>
>>>> On Monday, September 1, 2014 11:39:55 AM UTC+2, Martin Kleinschrodt
>>>> wrote:
>>>>>
>>>>> After some more experimentation, I narrowed it down a little more: It
>>>>> seems the issue was introduced between versions 0.3.4 and 0.3.5. Also, it
>>>>> seems that the problem is inside the polymer package, *not* the
>>>>> platform layer as one might expect (I reproduced the crashes with
>>>>> Polymer/polymer#0.3.5 + Polymer/platform#0.3.2). Then again, apparently
>>>>> TemplateBinding and polymer-expressions are contained in the polymer
>>>>> package, so maybe not that surprising after all. So it seems the culprit 
>>>>> is
>>>>> hiding in one of these changesets:
>>>>>
>>>>> https://github.com/Polymer/polymer-dev/compare/0.3.4...0.3.5
>>>>> https://github.com/polymer/TemplateBinding/compare/0.3.4...0.3.5
>>>>> https://github.com/polymer/polymer-expressions/compare/0.3.4...0.3.5
>>>>> https://github.com/Polymer/polymer-gestures/compare/0.3.4...0.3.5
>>>>>
>>>>> Anything else I missed that is part of the polymer package?
>>>>>
>>>>> On Monday, September 1, 2014 10:56:18 AM UTC+2, Martin Kleinschrodt
>>>>> wrote:
>>>>>>
>>>>>> Ever since updating from v0.3.2, my Polymer app crashes all the time.
>>>>>> The problem is extremely hard to debug since I don't get any error 
>>>>>> messages
>>>>>> in the browser console but instead the app crashes completely. This 
>>>>>> happens
>>>>>> both in mobile safari and in the web view. The only clue that I have is
>>>>>> this log from my cordova app:
>>>>>>
>>>>>>   Process 15971 stopped
>>>>>> * thread #6: tid = 0x5783c3, 0x36ae071a
>>>>>> WebCore`WebCore::Document::touchEventsChangedTimerFired(WebCore::Timer<WebCore::Document>*)
>>>>>> + 254, name = 'WebThread', stop reason = EXC_BAD_ACCESS (code=1,
>>>>>> address=0xfffffff9)
>>>>>>     frame #0: 0x36ae071a
>>>>>> WebCore`WebCore::Document::touchEventsChangedTimerFired(WebCore::Timer<WebCore::Document>*)
>>>>>> + 254
>>>>>>  
>>>>>> WebCore`WebCore::Document::touchEventsChangedTimerFired(WebCore::Timer<WebCore::Document>*)
>>>>>> + 254:
>>>>>> -> 0x36ae071a:  ldr.w  r8, [r0]
>>>>>>    0x36ae071e:  cmp.w  r8, #0x0
>>>>>>    0x36ae0722:  beq    0x36ae07e4                ;
>>>>>> WebCore::Document::touchEventsChangedTimerFired(WebCore::Timer<WebCore::Document>*)
>>>>>> + 456
>>>>>>    0x36ae0724:  movs   r0, #0x0
>>>>>>
>>>>>> I don't know enough about WebCore to get anything from this, but
>>>>>> based on when the crashes occur (usually after updating a model) I 
>>>>>> suspect
>>>>>> that the problem has something to do with either polymer-expressions or
>>>>>> TemplateBinding.
>>>>>>
>>>>>> I'm currently using Polymer 0.4.0 but I think the problem exists
>>>>>> since 0.3.3.
>>>>>>
>>>>>> If you want to reproduce the problem yourself, go ahead and open
>>>>>> http://maklesoft.github.io/padlock on your iPhone (or iOS simulator)
>>>>>> and follow these steps:
>>>>>>
>>>>>> - Choose a master password (you can simply use an empty one)
>>>>>> - Create a record
>>>>>> - Go back
>>>>>> - Create another record
>>>>>> - Switch back and forth between the two records (tap a record to
>>>>>> select it), at some point the app will crash (probably sooner rather than
>>>>>> later)
>>>>>>
>>>>>> Other ways to reproduce the crash
>>>>>>
>>>>>> - Delete a record
>>>>>> - Add a field to a record
>>>>>> - Add a category to a record
>>>>>>
>>>>>> I wish I could offer more information, but at this point I am pretty
>>>>>> much clueless. If you want to dig through the code, you can find it at
>>>>>>
>>>>>> https://github.com/maklesoft/padlock
>>>>>>
>>>>>> I probably don't have to mention that this is a show stopper for me.
>>>>>> I wonder if anybody is having similar problems?
>>>>>>
>>>>>      Follow Polymer on Google+: plus.google.com/107187849809354688692
>>> ---
>>> You received this message because you are subscribed to the Google
>>> Groups "Polymer" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to [email protected].
>>>  To view this discussion on the web visit
>>> https://groups.google.com/d/msgid/polymer-dev/8b2b71eb-2508-4aea-a7c2-de1ad34bec25%40googlegroups.com
>>> <https://groups.google.com/d/msgid/polymer-dev/8b2b71eb-2508-4aea-a7c2-de1ad34bec25%40googlegroups.com?utm_medium=email&utm_source=footer>.
>>>
>>>
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>
>>
> Follow Polymer on Google+: plus.google.com/107187849809354688692
> ---
> You received this message because you are subscribed to a topic in the
> Google Groups "Polymer" group.
> To unsubscribe from this topic, visit
> https://groups.google.com/d/topic/polymer-dev/YE0U38xqXig/unsubscribe.
> To unsubscribe from this group and all its topics, send an email to
> [email protected].
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/polymer-dev/CAAUAVAi2HB-eveg06Q0Be3VAEgtgPZdX790%2BOSyzz2sowrCx2w%40mail.gmail.com
> <https://groups.google.com/d/msgid/polymer-dev/CAAUAVAi2HB-eveg06Q0Be3VAEgtgPZdX790%2BOSyzz2sowrCx2w%40mail.gmail.com?utm_medium=email&utm_source=footer>
> .
>
> For more options, visit https://groups.google.com/d/optout.
>
>

Follow Polymer on Google+: plus.google.com/107187849809354688692
--- 
You received this message because you are subscribed to the Google Groups 
"Polymer" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/polymer-dev/CAAUAVAh-Pjp2qjYO-K15zPwdgoozhxHb15HMO7c1Q%2B1t-Dt3pg%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to