Bug 117933 – Crash in ASSERTION FAILED: !m_resource->errorOccurred() when loading mhtml archive file

• Component: Page Loading
• Status: UNCONFIRMED

[Santosh Mahto]

Program received signal SIGSEGV, Segmentation fault.
0xb737aa76 in WTFCrash () at /home/test/open_source/wk3/WebKit/WebKit/Source/WTF/wtf/Assertions.cpp:339
339	    *(int *)(uintptr_t)0xbbadbeef = 0;
(gdb) bt
#0  0xb737aa76 in WTFCrash () at /home/test/open_source/wk3/WebKit/WebKit/Source/WTF/wtf/Assertions.cpp:339
#1  0xb30da837 in WebCore::SubresourceLoader::didFinishLoading (this=0x9b4fc80, finishTime=0) at /home/test/open_source/wk3/WebKit/WebKit/Source/WebCore/loader/SubresourceLoader.cpp:269
#2  0xb30913b4 in WebCore::DocumentLoader::substituteResourceDeliveryTimerFired (this=0x9b3d610) at /home/test/open_source/wk3/WebKit/WebKit/Source/WebCore/loader/DocumentLoader.cpp:1120
#3  0xb309c3c0 in WebCore::Timer<WebCore::DocumentLoader>::fired (this=0x9b3db6c) at /home/test/open_source/wk3/WebKit/WebKit/Source/WebCore/platform/Timer.h:113
#4  0xb325c61b in WebCore::ThreadTimers::sharedTimerFiredInternal (this=0x91e6738) at /home/test/open_source/wk3/WebKit/WebKit/Source/WebCore/platform/ThreadTimers.cpp:129
#5  0xb325c533 in WebCore::ThreadTimers::sharedTimerFired () at /home/test/open_source/wk3/WebKit/WebKit/Source/WebCore/platform/ThreadTimers.cpp:105
#6  0xb3e0b027 in WebCore::timerEvent () at /home/test/open_source/wk3/WebKit/WebKit/Source/WebCore/platform/efl/SharedTimerEfl.cpp:52



The crash is happening because in  funtion when The cachedImage decoding fails it still calls didFinishLoading
DocumentLoader::substituteResourceDeliveryTimerFired(WebCo
{

loader->didReceiveData(data->data(), data->size(), data->size(), DataPayloadWholeResource)--> can DecodeFail  
......
.......
loader->didFinishLoading(0); ---> calling this without checking for errorOccured()
}

[Santosh Mahto]

Created attachment 205291 [details]
Patch

[Alexey Proskuryakov]

Is this covered by regression tests? Which tests asserts?

[Chris Dumez]

Comment on attachment 205291 [details]
Patch

WebCore fixes should have a layout test.

[Santosh Mahto]

Created attachment 206335 [details]
Crash on Load test case

( I checked in efl port)
step to reproduce crash:
1. load attached .mht file 

OR

1. ctrl + s(save)  http://m.news.nate.com/photoSlide/view?clstId=434328&section=sisa 
and load the saved mht file.

Attached mht is just saved file for this page
http://m.news.nate.com/photoSlide/view?clstId=434328&section=sisa 

@chris : i will check if i can add layout test

[Santosh Mahto]

Created attachment 206651 [details]
Patch

[Alexey Proskuryakov]

Comment on attachment 206651 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=206651&action=review

> Source/WebCore/loader/DocumentLoader.cpp:1122
> +                loader->cancel();

I'm not sure if canceling is an appropriate response here. We are failing the load, shouldn't it be handled as a failure?

[Alexey Proskuryakov]

Also, can you make this happen with ApplicationCache? MHTML is disabled on many platforms, and having a cross-platform regression test will be a better protection against future regressions.

[Michael Catanzaro]

Comment on attachment 206651 [details]
Patch

Clearing the r? flag to get this out of request queue. If you still want to work on this, please do respond to Alexey's latest feedback. Thanks!