1281583 - [x64] Crash in CDecodeBuffer::`scalar deleting destructor''

Status: RESOLVED INCOMPLETE

(External Software Affecting Firefox Graveyard :: Flash (Adobe), defect, P3)

Description

[YF (Yang)]

This bug was filed from the Socorro interface and is 
report bp-9626668e-b200-4919-a45a-6971d2160622.
=============================================================

STR:
Open the http://www.bilibili.com/video/av5033867/ and wait for the player to loaded.

Actual results:
Flash Player always crashes in Fx47 & Fx48.0b1, only for x64 firefox, no crashes in x86 versions. no browser crashes. not e10s enabled.


The regression-window:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=9c43645228d84d1eefcd03cb2944244253282f99&tochange=b71d96329d7aa31c01b13319707aa96587051f0b

Comment 1

[Benjamin Smedberg]

I couldn't reproduce this locally, but that doesn't mean much because this is going to be a crash that's dependent on the machine graphics config and accelerated graphics settings. 

The crash appears to be entirely within the D3D11 code. The system threw an exception and was unwinding to clean up from it. While deleting stuff during the unwind, it's running a CDecodeBuffer destructor.
It's calling a virtual destructor on some internal member pointer of the CDecodeBuffer (this+0x18) except the vtable pointer is NULL so we crash.

I don't know whether the problem is that the system threw an exception, or that the unwinding code is broken.

In either case Michelle can you spend time to see if we can come up with a reliable STR/config so that we can hand this to Adobe engineering. It's unlikely to be a bug on the Firefox side or even something we can effectively work around.

Comment 2

[Michelle Funches - QA]

Complete: could not reproduce.
Full test information is detailed in email "Bug 1281583 Adobe and Bilibili Graphics"

Comment 3

[Michelle Funches - QA]

Attachment: Adobe_Bilibili Graphics Test Report

Attaching report that details the environment and steps applied.

Comment 4

[Benjamin Smedberg]

Unfortunately without good STR it's going to be hard to do much about this. I'm going to leave it in the backlog for now, in hopes of finding more people who see this and finding a pattern.

Comment 5

[YF (Yang)]

(In reply to Michelle Funches - QA from comment #3)
> Created attachment 8764622 [details]
> Adobe_Bilibili Graphics Test Report
> 
> Attaching report that details the environment and steps applied.

I see the "User Agent 	Mozilla/5.0 (Windows NT 10.0; WOW64; rv:47.0) Gecko/20100101 Firefox/47.0" in this report, but this bug should use Firefox x64 version to reproduce.

Comment 6

[Michelle Funches - QA]

Confirmation: Yes Firefox x64 was used in the testing process.
Apologies for misleading info. May be a typo or bad cut n paste.

User Agent 	Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:47.0) Gecko/20100101 Firefox/47.0
OS 	        Windows_NT 10.0 x86-64

Comment 7

[YF (Yang)]

I see new crash signatures with this page in Fx52b4 x64 / x86.

Comment 8

[u597032]

Adobe Flash is no longer supported.