1028424 - Crashes at libsystem_kernel.dylib@0x1637a (pthread_kill | abort) on Yosemite (OS X 10.10)

Status: RESOLVED INVALID

(Core :: General, defect)

Description

[Steven Michaud [:smichaud] (Retired)]

https://crash-stats.mozilla.com/query/?product=Firefox&version=ALL%3AALL&platform=mac&range_value=4&range_unit=weeks&date=06%2F20%2F2014+22%3A00%3A00&query_search=signature&query_type=contains&query=libsystem_kernel.dylib%400x1637a&reason=&release_channels=&build_id=&process_type=any&hang_type=any

These happen on all branches, and in old code.  But there are many more of them on the 33 branch, where they're actually the #1 Mac topcrasher.  This is probably because our very small number of Yosemite users mostly use trunk nightlies.

I assume this is something stupid, and a bad interaction between Firefox and Yosemite.  But I don't yet know whose bug it is, and in fact I don't yet have Yosemite.

Josiah, I'm pretty sure you have it.  Have you seen these, and can you reproduce them?

Comment 1

[Josiah Bruner [:jsbruner]]

I have actually not encountered any crashes on 10.10 myself (with the 33a Nightly).

My Yosemite machine is an older desktop though, and I therefore lack a Retina display or the dual graphics cards like the newer models. And since these crashes appear graphics related, I wonder if those components are relevant?

Comment 2

[Josiah Bruner [:jsbruner]]

Oh, I forgot to mention, I could also not reproduce these crashes using any of the steps/causes mentioned in the comments of those crash reports.

Comment 3

[Steven Michaud [:smichaud] (Retired)]

I just installed Yosemite (DP2), and haven't been able to reproduce these crashes either.

I installed it on a roughly two-year old Retina MacBook Pro (15" MacBookPro10,1).

Comment 4

[Steven Michaud [:smichaud] (Retired)]

Since this is a (relatively) huge topcrasher, I'll try to get to it later this week.

Comment 5

[Steven Michaud [:smichaud] (Retired)]

Looking at these more closely, I find lots of different kinds of stacks.

I also find that the NS_CompareVersions() stack entries are almost certainly spurious.

I'll keep digging into this ... but there may be several different bugs here :-(

Comment 6

[Steven Michaud [:smichaud] (Retired)]

All these reports share the following top three lines (translated using atos):

__pthread_kill (in libsystem_kernel.dylib) + 10
pthread_kill (in libsystem_pthread.dylib) + 90
abort (in libsystem_c.dylib) + 129

That just means something in system code is choking and aborting the process -- which of course isn't the bug itself, but a symptom of it.

Because Socorro doesn't let us search on anything but the top line of a crash stack (the so-called "signature"), I'll need to pore through these reports by hand to find out how many different kinds of crashes share these top three lines (and how many different kinds of bug we actually have).

Comment 7

[Steven Michaud [:smichaud] (Retired)]

I just noticed that very few of these are on the current version of Yosemite (14A238x) -- DP2 plus its latest update.  So I'm going to wait a while to see if Apple has already mitigated these crashes.

Comment 8

[u408661]

Attachment: backtrace.txt

I can reproduce this crash reliably by trying to download the patch from bug 1030203 (just right-click on the name, and select "Save Link As..." for the fun). This is on Yosemite build 14A261i. I also reproduced just trying to attach this backtrace as an attachment (I'm doing this one in Safari), so it seems (at least in my case) to be related to bringing up the file picker.

Comment 9

[Steven Michaud [:smichaud] (Retired)]

OS X has a long history of file picker bugs, going back at least to OS X 10.7.  These were/are almost certainly Apple bugs, but they sometimes effected Firefox badly (they were sometimes Mac topcrashers).

Now and then, over the years, I've tried to reproduce them.  I've never been successful.

So if I had a copy of Yosemite DP1 (which I think corresponds to build 14A261i), or could get hold of one, I'd install it just to see if, finally, I'd have a build with which I *can* reproduce one of these Apple file picker bugs -- if only to understand what causes them.  But Yosemite DP1 isn't (any longer) available from Apple's Mac Developer Center (https://developer.apple.com/membercenter/), which is the only place I know of where one can get a copy.

I don't know why DP2's build number (14A238x) is higher that what appears to be DP1's build number (14A261i).  Can you shed any light on that?

And for the record, where did you get your copy of Yosemite?

Comment 10

[Steven Michaud [:smichaud] (Retired)]

And no, your STR from comment #8 doesn't work for me.  I just tried it in a recent m-c nightly.  I saw no problems (and certainly no crashes).

Comment 11

[Steven Michaud [:smichaud] (Retired)]

> I don't know why DP2's build number (14A238x) is higher that what
> appears to be DP1's build number (14A261i).  Can you shed any light
> on that?

I don't know why DP2's build number (14A238x) is *lower* than what
appears to be DP1's build number (14A261i).

Comment 12

[Steven Michaud [:smichaud] (Retired)]

Actually DP2's build number does appear to be 14A261i.  But after installing "Update 1.0" from the App Store I ended up with build number 14A238x.

Comment 13

[u408661]

Odd that you ended up with that build number, as I have also installed Update 1.0 from the app store, and still show 14A261i. For the record, my original install was via the developer center.

My crashes are on the most recent nightly. I'm using a MBP 8,2.

Comment 14

[Nomis101]

(In reply to Steven Michaud from comment #11)
> > I don't know why DP2's build number (14A238x) is higher that what
> > appears to be DP1's build number (14A261i).  Can you shed any light
> > on that?
> 
> I don't know why DP2's build number (14A238x) is *lower* than what
> appears to be DP1's build number (14A261i).

I think its the other way. My DP2 is 14A261i.

Comment 15

[Steven Michaud [:smichaud] (Retired)]

Actually I just was prompted to *re*-install Update 1.0, and got a failure message after restarting.  This probably also happened earlier without me noticing.  I'm pretty sure that both times I did the update, I updated iTunes and the OS at the same time.  Now I'm trying to update just the OS.

I have a roughly two-year-old Retina MacBook Pro (10,1).

Comment 16

[Steven Michaud [:smichaud] (Retired)]

Installing iTunes and Update 1.0 did the trick.  Now I have build 14A261i, like the rest of you.  Sorry for the confusion.

But no, the STR from comment #8 still doesn't crash for me.

Nicholas, I see that your crash stack includes references to "QuickLook" stuff.  I wonder where that comes from?  What's your default "view" format in Finder?  (I have "View" "as List").

Comment 17

[u408661]

I'm configured for "View"->"as Columns"

Comment 18

[Steven Michaud [:smichaud] (Retired)]

Nicholas, do you have any OS extensions (or whatever) that change how file pickers are displayed?

Do you have anything installed in /Library/Input Methods/ (a notorious source of trouble)?

Did you do a clean install of Yosemite, or did you upgrade from a previous version of OS X?

(I changed to View -> as Columns, but this didn't make any difference.  I still don't crash.)

Comment 19

[Steven Michaud [:smichaud] (Retired)]

> Installing iTunes and Update 1.0 did the trick.

Installing iTunes and Update 1.0 *separately* did the trick.

Comment 20

[u408661]

(In reply to Steven Michaud from comment #18)
> Nicholas, do you have any OS extensions (or whatever) that change how file
> pickers are displayed?

Not unless someone slipped one on this computer when I wasn't looking :)

> Do you have anything installed in /Library/Input Methods/ (a notorious
> source of trouble)?

Nope, it's empty.

> Did you do a clean install of Yosemite, or did you upgrade from a previous
> version of OS X?

This was an upgrade from whatever version of 10.9 was current as of Monday (when I upgraded). 10.9 was most likely (I can't remember any more) also an upgrade from 10.8.whatever way back when (though I never had any betas previously, Yosemite is my first OS X beta).

> (I changed to View -> as Columns, but this didn't make any difference.  I
> still don't crash.)

And I tried changing to View->as List, and it also didn't make any difference, I still crashed :)

Comment 21

[Steven Michaud [:smichaud] (Retired)]

Thanks, Nicholas, for trying.  But the reason you crash and I don't is probably too complex for us to figure out.  If you ever do find out, I'd be *very* interested to know.  As I said, I've been trying to reproduce these file picker crashes for years.

One last suggestion:  Try rebooting your computer and see if this makes the crashes go away, at least temporarily.

Comment 22

[u408661]

Sadly, a reboot didn't fix the crashes at all, so this remains a mystery. I'm not sure what else to do at this point to figure it out, although I definitely downloaded resources between my original install of Yosemite and today (whether they were before or after I installed Update 1.0, I'm honestly not certain... all my downloads were on Tuesday, the same day I installed the update, with the exception of one, which never involved the file picker, as it was a PDF I opened directly in Preview). If I manage to get any more information out of this, I'll be sure to update the bug.

Comment 23

[Werner E. Goedecke R.]

I can reliably reproduce the crash by doing the following:

Click "Tools" > "Web Developer" > "App Manager" > "Apps" > "Add Packaged App"

These are a couple crashes:

Report ID 	Date Submitted
bp-cf67ebfc-1cdb-4202-a70a-9b68a2140629	29/06/14	06:16 a.m.
bp-04ce7cdb-af09-46df-9f1d-f50dd2140629	29/06/14	06:14 a.m.

Comment 24

[Steven Michaud [:smichaud] (Retired)]

(In reply to comment #23)

Please post a link to examples of "packaged apps" that I can play with.

Comment 25

[u408661]

FWIW, I just upgraded one of my macs to Yosemite release (same machine as I was using the developer previews on), and this bug appears to have been fixed on Apple's side. I can "Save As" just fine. Additionally, the steps referenced in comment 23 just bring up the file picker to choose a packaged app to test, so I imagine it was the same issue as the "Save As" bug I was seeing.

My guess is we can close this as either WORKSFORME or INVALID.

Comment 26

[Steven Michaud [:smichaud] (Retired)]

There are still some of these in the Yosemite release:

https://crash-stats.mozilla.com/report/list?signature=libsystem_kernel.dylib%400x16282&product=Firefox&query_type=contains&range_unit=weeks&process_type=any&platform=mac&hang_type=any&date=2014-10-22+17%3A00%3A00&range_value=4#reports

But they're no longer anywhere near a topcrasher.

So it's premature to resolve this bug (as WORKSFORME or anything else).  But it's not worth much attention until and unless it becomes a topcrasher again.

Comment 27

[Lawrence Mandel [:lmandel] (use needinfo)]

Given comment26, I'm dropping the tracking flags. Please renom if this becomes more serious again.

Comment 28

[Simon Howes]

I think this may be related to this bug. Click "Test your browser" and OS X GPU driver crashes.

http://ie.microsoft.com/testdrive/graphics/webglstresstest/


Here is OS X crash log if this is of any help:

Wed Jan 21 21:39:28 2015

Event:               GPU Reset
Data/Time:           Wed Jan 21 21:39:28 2015
Application:         firefox
Path:                
OS Version:          Mac OS X Version 10.10.1 (Build 14B25)
Graphics Hardware:   NVIDIA GeForce GT 750M
Signature:           d

Report Data:

NVDA(OpenGL): Channel exception! Exception type = 0xd Graphics Engine Error (GR Error 5)
Channel Info: [14, 0x12, 0xe, 0x1bb1]
Version Info: [com.apple.GeForce, 10.0.0, 0x746c6943485b52, 18831271, 310.41.05f01, 1]

Resource Manager Info:
 4443564e 00000158 ca5e870e f997d2b4 00000001 00000014 d3793533 46d3a4a6
 4614f297 e71edccf 00088301 00000124 12ea3f0a 080c0a3c 18121000 ace92001
 2a0292d7 1001080b 02808288 01800222 01080e2a 81899010 84052202 08808080
 01080f2a 81899410 db062202 0180808d f2530a01 250a5012 02c19708 80808010
 20001801 30002800 1f883800 809f8840 80480880 504080c0 60125800 080a2200
 8e801001 01220280 080b2200 8a841001 02220280 0e221c82 84100108 22028180
 80828005 030a0884 0a00149a 1f138222 2d100008 f7200118 e1b02806 50004804
 60015820 78007064 01019000 0a000198 00138a03 13923d0a 24380a3a 0e000000
 01000000 490000e0 0100000f 49000000 0000000b 47000000 ff000304 ff000000
 ff000000 ff000000 ff000000 ff000000 0a000000 1c13c21f 00100008 8480a818
 12200bf7 03300d28 edd58038 4010d08f 03480280 4443564e

Accelerator Event History:
 0a0808001a04080010010a0808001a04080210010a2a0800122608bc80021080e0c9d0
 84f0ffffff0118f181c08eec8180092082808080e02b28b2b78080e0010a230800121f
 08c080021080e0c9d084f0ffffff0118f181c08eec818009208182800828000a080800
 1a04080210000a0808001a0408001000

Comment 29

[Steven Michaud [:smichaud] (Retired)]

Simon, your crash report is almost certainly unrelated.  It doesn't concern a filepicker crash.

And why is it truncated?  Is what you posted all there was to it?

Comment 30

[Steven Michaud [:smichaud] (Retired)]

But I also crash (probably a different crash) with that test.  I'll check if we already have a bug on the test, and if not I'll open one.

Comment 31

[Simon Howes]

Yes, that is all of the log.

I come to this bug as the GPU driver crashes, and then so does the browser. This is the browser's crash report:

https://crash-stats.mozilla.com/report/index/906e3b50-ac4a-4366-88c5-1fcde2150122

It links to this bug as a related bug. Hence why I came here.

Comment 32

[Steven Michaud [:smichaud] (Retired)]

> It links to this bug as a related bug. Hence why I came here.

I don't know why.  This bug is completely unrelated to your crash report.

I've opened bug 1124427 on the crashes caused by Microsoft's WebGL Stress Test.

Comment 33

[BugBot [:suhaib / :marco/ :calixte]]

Crash volume for signature 'libsystem_kernel.dylib@0x16282':
 - nightly (version 50): 0 crash from 2016-06-06.
 - aurora  (version 49): 0 crash from 2016-06-07.
 - beta    (version 48): 1 crash from 2016-06-06.
 - release (version 47): 277 crashes from 2016-05-31.
 - esr     (version 45): 25 crashes from 2016-04-07.

Crash volume on the last weeks:
             Week N-1   Week N-2   Week N-3   Week N-4   Week N-5   Week N-6   Week N-7
 - nightly          0          0          0          0          0          0          0
 - aurora           0          0          0          0          0          0          0
 - beta             0          0          0          0          1          0          0
 - release         54         41         32         46         43         33          9
 - esr              4          2          6          4          3          2          2

Affected platform: Mac OS X

Comment 34

[BugBot [:suhaib / :marco/ :calixte]]

The bug assignee didn't login in Bugzilla in the last months and this bug has severity 'critical'.
:overholt, could you have a look please?
For more information, please visit auto_nag documentation.

Comment 35

[Andrew Overholt [:overholt]]

Should we keep this open, Haik? Should we move it to Widget:Cocoa?

Comment 36

[Haik Aftandilian [:haik]]

I don't think we need to keep this open. A user in comment 25 reported the problem stopped occurring for them and it was likely fixed in macOS. Today macOS 10.10 (Yosemite) is no longer supported with our minimum macOS version being 10.12.