1324424 - Crash reporter crashes itself on Mac OS 10.12 with EXC_BAD_ACCESS (SIGSEGV)

Status: RESOLVED DUPLICATE of bug 1320048

(Toolkit :: Crash Reporting, defect)

Description

[Hanno Schlichting]

Attachment: crashreporter_2016-12-02-092742_xahho-book.crash

I have one of the new 15" Macbook Pro's with a touch bar, running Mac OS 10.12 (both .1 and .2). On this machine Firefox 51 beta itself crashes reliably every time the OS wakes up from sleep.

In addition crash reporter crashes about once or twice per day (also each time seconds after waking up), so no Firefox crash report is captured by it. Those crashes are captured and send to Apple by the OS crash handling.

I've attached a couple crash reports with varying exception codes. For example:

Crashed Thread:        5  Dispatch queue: com.apple.root.user-interactive-qos.overcommit

Exception Type:        EXC_BAD_ACCESS (SIGSEGV)
Exception Codes:       KERN_INVALID_ADDRESS at 0x000007fa399ca1f0
Exception Note:        EXC_CORPSE_NOTIFY


Crashed Thread:        0  Dispatch queue: com.apple.main-thread

Exception Type:        EXC_BAD_ACCESS (SIGSEGV)
Exception Codes:       EXC_I386_GPFLT
Exception Note:        EXC_CORPSE_NOTIFY


Crashed Thread:        4  Dispatch queue: com.apple.touchbar.agent

Exception Type:        EXC_BAD_ACCESS (SIGSEGV)
Exception Codes:       KERN_INVALID_ADDRESS at 0x0000000000000018
Exception Note:        EXC_CORPSE_NOTIFY


Crashed Thread:        0  Dispatch queue: com.apple.main-thread

Exception Type:        EXC_BAD_ACCESS (SIGSEGV)
Exception Codes:       KERN_INVALID_ADDRESS at 0x0000000000000020
Exception Note:        EXC_CORPSE_NOTIFY

Comment 1

[Hanno Schlichting]

Attachment: crashreporter_2016-12-15-064821_xahho-book.crash

Comment 2

[Hanno Schlichting]

Attachment: crashreporter_2016-12-18-104543_xahho-book.crash

Comment 3

[Hanno Schlichting]

Attachment: crashreporter_2016-12-19-130615_xahho-book.crash

Comment 4

[Gabriele Svelto [:gsvelto]]

Those crashes all look different, this is quite weird. Does the crashreporter crash after having shown its UI or does it crash right away after Firefox' crash?

Comment 5

[Hanno Schlichting]

I don't see the crash reporter UI coming up, but only see Apple's built-in crash handler. It looks like the crashes happen 1-5 seconds after the laptop wakes up from sleep. During that time as a user I just see the OS-level password unlock screen and than use touch-id to unlock the laptop. So by the time I've unlocked the machine, it's too late and the crash reporter has already crashed.

My guess is that the introduction of the new "Dispatch queue: com.apple.touchbar.agent" thread in the application has caused this and might now trigger some threading related issue that wasn't there before. I'm guessing the OS magically introduces that new thread without any changes to the app code.

But I can't tell whether the issue is with the way that injection is done or if this exposed an actual app level problem that wasn't apparent before with fewer threads around.

Comment 6

[Hanno Schlichting]

Googling around a bit for the same kind of problem I've stumbled upon various reports which hint at this being a wider spread problem. One report claims this happens more frequently with apps linked against older SDK versions:

- http://indiestack.com/2016/12/touch-bar-crashers/
- http://www.openradar.me/radar?id=5063875711991808
- http://mjtsai.com/blog/2016/12/16/touch-bar-crash-protection/
- https://github.com/racket/racket/issues/1520

Comment 7

[(not currently active) Ted Mielczarek]

This sounds like it might be fundamentally the same crash as bug 1320048.

Comment 8

[Wayne Mery (:wsmwk)]

Hanno, does also occur for version 52? (aurora branch)
Do you happen to also run Thunderbird?

Comment 9

[Wayne Mery (:wsmwk)]

As of this morning, 3 Thunderbird 51.0b1 beta users who crashed with bug 1286613 have tried 52.0a2, and report they also crash with TB 52.0a2 AND crash reporter is crashing.  

I suspect this bug may be keeping many Thunderbird Mac users away from aurora - because based on crash-stats I'm seeing way more 51.0a2 crashes than 52.0a2 crashes, even though 52.0a2 has been out for a couple months

Comment 10

[Hanno Schlichting]

I'm not using Thunderbird and haven't installed Firefox Aurora. Based on bug 1320048, comment 50 it looks to me like the current plan is to wait for Apple to fix the issue in Mac OS 10.12.3 (in beta now). There seems to be no easy workaround on the Firefox / crashreporter side.

So I'd be inclined to close this as a duplicate of bug 1320048.

The long term fix seems to be to upgrade the build infrastructure to use a later Mac OS SDK (bug 1324892).