Google admits Chrome bug responsible for crashing MacBook Airs - Page 2

Chrome was my default browser, but I gave up using it a few months ago because it was causing my 2009 MacBook Pro to freeze every couple of days or so (running current version of OS X). Now I've gone back to using Firefox it never freezes. I occasionally start Chrome just to see if there''s been a software update, because I dare not use it as it is.

Quote:

Originally Posted by hill60 

 

Actually Google are fans of Safari seeing as how that's what Chrome is based on as can be seen in the open source license agreements of Chrome and Android.

 

Try putting things the right way around the next time.

Chrome isn't based on Safari, it's based on WebKit. WebKit is an Apple-developed open source rendering engine used by practically every web-browser apart from IE and Firefox.

Ok.

Why is this a major, front-page news item?

Quote:

Originally Posted by myapplelove 

All joking and google bashing aside, it's shocking how this can happen to os x in 2012, a gfx driver crash hasn't brought down windows since vista in what 2007? The graphics stack should simply restart without bringing down the whole os. There 's a problem here that needs to be fixed in the graphics driver, but the underlying problem which I can assume wont be fixed in ml, of the os allowing this to happen is what's really troubling here.

My Windows 7 install freaks out all the time on my iMac when trying to shove games down its neck. Sometimes the AMD graphics driver restarts, once I got a blue screen and twice already I've had the drivers crash but they didn't recover; so the monitor never came back on and the computer stopped responding to input (even incoming network connections).

Quote:

Originally Posted by mausz 

 

But windows 7 did not crash at the kernel level.

 

If a program crashes, only the program should crash. If a program crashes a driver, only the driver should crash (and restart). If you can lockup the kernel using a driver this will lead to potential hacks which can run at kernel level.

If just the application fails then good for Windows! If, however, (and you see this quite often with Windows) the entire OS freezes and you can't even move the mouse, then your computer at the hardware level has freaked out completely and halted everything. Some computers signal this error by giving one, low beep tone from the internal speaker. If you don't hear the beep tone then listen for hard disk activity - no activity means halted computer. Its a Kernel Panic++

I'm not saying the Mac doesn't do that (I've had it happen personally), but its definitely a more common occurrence in Windows than on OSX. Chrome on my Work computer has caused this to happen already, so I've uninstalled it and put Chromium (the open source version) on here instead. A few version numbers ahead and running like a dream.

Quote:

Originally Posted by myapplelove 


Sure, the problem is though that the os shouldn't kerrnel panic because of that and crash, that's the whole pint of making a robust os in 2012. It should safeguard itself even from the most malicious code. What's also conveniently ommited here is that it's not the intel chip, it's the drivers for the intel chip, there's no problem on the hardware level with the intel chip. This case is revealing a much deeper vulnerability of os x.

Please. I just had a blue screen on my Win 7 x64 box yesterday. No OS can be protected as long as ring 0 access is allowed and that's where these drivers run. But that's what's necessary to provide performance. 

Quote:

Originally Posted by MacBook Pro 


His statement is a blatant mistruth. One need only type "graphics blue screen of death" into any search engine for an absurd number of complaints about bug checks ("Blue Screen of Death") in Microsoft Windows caused by graphics drivers. In fact, his statement demonstrates a complete lack of understanding the issue.
Considering the features and performance of Google Chrome for iPhone versus Apple Safari for iPhone there is little to no point in using Google Chrome:
iOS 6 beta 2 on iPhone 4S:
Rightware Browsermark Chrome for iPhone
1. 48158
2. 49222
3. 48777
Rightware Browsermark Safari for iPhone
1. 101458
2. 107252
3. 105743
Sunspider JavaScript 0.9.1 Google Chrome for iPhone
1. 6829.8ms
2. 6836.6ms
3. 6834.7ms
Sunspider JavaScript 0.9.1 Apple Safari for iPhone
1. 1760.5ms
2. 1752.5ms
3. 1732.5ms
Apple Safari features that Google Chrome lacks; however, in all fairness Google may require one to login to Google for some features which I am not foolish enough to do:
No synchronized tabs
No synchronized bookmarks
No offline reading list
No ability to add to Home Screen
No Fraudulent website warnings
No Facebook integration (likely to change in update prior to or once iOS 6 is released)
No Message integration (likely to change in update prior to or once iOS 6 is released)
No Printing (likely to change in update prior to or once iOS 6 is released)
No Twitter integration (likely to change in update prior to or once iOS 6 is released)

Not the most informed post I have to say. Chrome on iOS is slower than Safari as *Apple* does not allow programs which use uiwebview (browsers cannot have their own rendering engine on iOS) to use the Nitro javascript engine.

Chrome does have synchronization of tabs and bookmarks (and in my experience does a better job than iCloud). Of course you need to sign in to use them, how else could that possibly work?! The other features, I'll grant you that.

Quote:

Originally Posted by hill60 

 

Tried it once, that was enough, kept crashing due to sneaking in Adobe Flash crap.

 

Safari works fine, so does Firefox.

 

My MacBook remains Chrome free, don't need it, don't want it, don't care how popular it is.

I was having trouble last night with my 2010 Air ramping up the fans and using 90% of the CPU. Traced it to Chrome. I do not trust that browser any further than I can spit.

Quote:

Originally Posted by jragosta 


Exactly. Graphics drivers almost have to be granted kernel access and it's impossible to prevent a bad driver from having serious consequence. Well, you could, but you'd have to put the drivers in user space and then all the Apple haters would complain about the poor performance.
It's a tradeoff that almost every OS has made and relies on reliable drivers. The bad thing is that it is possible to bring the system down with an application bug. Rare, but it happens.
There are only two solutions:
1. Google fixes Chrome bug to prevent the problem.
2. Rewrite Mac OS X to put all the drivers in user space with all the expense, difficulty, and usability problems it would cause.
Obviously, the first is more practical.

This is true, but there are *certain* steps an OS can take to mitigate video driver crashes. Windows will attempt to restart a video driver before throwing a BSoD. http://support.microsoft.com/kb/2665946

Quote:

Originally Posted by sr2012 


Wow, if I had a dollar for every time Windows actually did that, since, say, Windows 2000, I'd be... broke.

It's only been a feature since Vista.

http://en.wikipedia.org/wiki/WDDM#Enhanced_fault-tolerance

Quote:

Originally Posted by sr2012 

Perhaps that's because the GPU overclocker "polls" the system very frequently to keep in touch with how the driver/GPU is doing?

No not the overclocker; what you describe is precisely what Windows does.

Quote:

Originally Posted by jragosta 


Those circumstances are very rare. The vast majority of graphics driver crashes (or, mostly, any driver running inside kernel space) don't give the system a chance to respond. Boom, you're dead.

"The biggest change in the WDDM [driver model since Vista] is that much of the graphics driver has been moved from kernel space to user space."

"In the unlikely event that an application or its UMD does something illegal and causes an error, only that single application will close, leaving the Windows Vista operating system unscathed, allowing the user to continue working."

Source: http://www.amd.com/us/Documents/ATIWDDMWhitepaperFinalV38.pdf

Quote:

Originally Posted by GTR 

 

"the specific combination of Intel HD 4000 chip + flash + Chrome."

 

Crash, Infection, Java, Flash, Plugin.

 

Always the same word combinations...

Software running in user mode should NEVER cause the OS to crash. At worst, the OS should freeze or kill an errant process.

Whilst the story here is trying to paint the cause of the crashes as being Google Chrome, the actual cause of the crashes is something within Mac OS X that Chrome has exposed (albeit via a bug within Chrome). It looks like the bug is within the HD4000 graphics driver, and it's likely that the code was supplied by Intel for Apple to integrate into Mac OS X.

Quote:

Originally Posted by sr2012 


OK fair enough, then I've never seen it from the time of Vista through Windows7 (and I've been through them all, 32bit, 64bit), except in rare times running GPU overclocking apps.
Could it have been working in the background unnoticed at other times? Perhaps, but I still did get BSODs every now and then.
For heaven's sake even in 7 and definitely in Vista, good luck Alt-Tabbing out of a fullscreen game. The moment you do that you pretty much lose all hope of actually returning to your game, unless you're willing to wait 5 minutes or more... Yes this is not a kernel panic as such but you see what I'm trying to illustrate.
So what I would say on the topic of Kernel Panics, is that Windows still has a big history stretching back decades, and arguably they have not addressed it sufficiently.

Absolutely. (Though I gotta say, the alt-tab behaviour you're describing is more likely due to poor coding in the game. Case in point: Valve released a patch to the source engine a while back making alt-tab in their games virtually instant) It is also absolutely true that the WDDM fault tolerance mechanism is far from perfect. I remember seeing a Microsoft document pointing out that despite WDDM 1.0, the sizable majority of BSoDs are still graphics driver related. But at least such a mechanism exists, and I can't seem to find any documentation of something similar in OS X.

Safari can't sync my bookmarks over several systems (including Linux at work). No use at all.

Quote:

Originally Posted by MacBook Pro 

Apple Safari features that Google Chrome lacks; however, in all fairness Google may require one to login to Google for some features which I am not foolish enough to do:
No synchronized tabs
No synchronized bookmarks
No offline reading list
No ability to add to Home Screen
No Fraudulent website warnings
No Facebook integration (likely to change in update prior to or once iOS 6 is released)
No Message integration (likely to change in update prior to or once iOS 6 is released)
No Printing (likely to change in update prior to or once iOS 6 is released)
No Twitter integration (likely to change in update prior to or once iOS 6 is released)

Chrome has progressed since the internal 0.1 build that must have been your last experience of the browser, I suggest you give it a try.

Who wants social network integration in their web browser anyway? If I wanted it I'd install an extension, but oddly enough I find the actual websites enough. Or dedicated apps.

Oops.  Yea, Chrome has a boo boo, but the thing that's worse is that an app was able to crash MacOS X.  Apple obviously still has some work to do with OS stability.  In this day and age, that should not have happened.

Quote:

Originally Posted by Damn_Its_Hot 


It is very possible to pass bad code at the level you are talking about (much closer to the silicon than other calls) if indeed they are trying to optimize the 4000. I can't see to much of and area where you could optimize code via user level calls.The API does allow for some stuff that you can hang yourself with if you are not fastidious in your resource/memory management. Very little else could do that other that a kext. A well written and thoroughly tested app should not see this.
There is not anything in the OS to keep poorly behaved code (badly written) from crashing an app. From what I read it simply crashed…
The moral of the story is stay away from third party browsers (and other Google and Adobe code -- they both have a track record of putting out good ideas without the best QA) until someone else has shaken it out.

Modern (since 1975) OSes are meant to shield users from bad programmers (and users from themselves).  The ability for a user level app to crash a system is an OS defect.  Period.   

As a passing AI reader... I read in the main article

" the machines were frequently crashing" 

meaning... crash.  reboot.    I don't know what your pronoun 'it'  means (Chrome?  Mac OS X?), but we may just be in violent agreement.

And maybe more to your point..  Should an OS* have levels other than 'user mode' that are code unlocked by a more rigorous QA process... say, the GPU primitives are only available to applications with Mac AppStore signed code.   Developers can self-sign for their own machines, but to run on a 'consumer' machine, it has to be signed by Apple to unlock the that 'low level' API.  Make the graphics primitive code call require passing the signature from Apple against the app, and then no GPU poking without at least Apple's QA team doing a once over

*In VMS parlance... it would have been a CMEXEC call to allow the app to run in Executive mode (VMS has 4 processor modes that could be assigned as privileges to a userID, Process, or Application: User, Supervisor (think 'operator'... mounting/dismounting media), Executive (typically direct driver calls), and kernel).  

Quote:

Originally Posted by jragosta 


Exactly. Graphics drivers almost have to be granted kernel access and it's impossible to prevent a bad driver from having serious consequence. Well, you could, but you'd have to put the drivers in user space and then all the Apple haters would complain about the poor performance.
It's a tradeoff that almost every OS has made and relies on reliable drivers. The bad thing is that it is possible to bring the system down with an application bug. Rare, but it happens.
There are only two solutions:
1. Google fixes Chrome bug to prevent the problem.
2. Rewrite Mac OS X to put all the drivers in user space with all the expense, difficulty, and usability problems it would cause.
Obviously, the first is more practical.

3.  Apple fix the graphics driver bug that the chrome code exercised.

The 3rd is most correct and fixes it for all programs.   OS stability should always trump performance (if you can optimize your GPU calls, great, but we are not letting the driver crash the system, by letting the 'user mode' call to the driver put data in that will put the device/driver/kernel in an inconsistent, unrecoverable state.

No application should ever be able to crash the operating system.  Bad design.

And that, folks, is why Adobe can't be trusted with access to low-level video hardware for Flash (or anything else).

And, yes, I did notice they finally gave up on Flash for Android installs.

Shantanu Narayen really should resign so Adobe can get back to the stuff that made them great, not the craptacular Macromedia stuff they bought.

Quote:

Originally Posted by Tallest Skil 


I question its popularity, as well. IE is still the standard in most of the world, even though 6 has thankfully now bitten the dust.

According to netmarketshare, IE6 still has more users than any (single) desktop edition of Safari, at 6.17% of the browser market. 

In what sense of the word has IE6 "bitten the dust"?

/hates IE6, I do

Quote:

Originally Posted by fredaroony 

Lot of people like obviously, it's currently the most used browser.

A lot of people are using it without even realizing it. I've spoke with a number of people for tech support that claim to be using "Internet Explorer", only to find out that they unknowingly switched to Chrome at some point without realizing it. A lot of people are drones when it comes to this stuff. And Google has been very aggressive about pushing Chrome onto people.

Quote:

Originally Posted by Curmudgeon 

No application should ever be able to crash the operating system.  Bad design.

That's an oversimplification. Hardware/software integration is complex stuff. In this case, Google's code works fine on other hardware, just not the _new_ hardware introduced by Apple. Now Google must change their code to accommodate that new hardware. This tail-chasing will always happen.

Quote:

Originally Posted by Hattig 

Safari can't sync my bookmarks over several systems (including Linux at work). No use at all.

 

 

Chrome has progressed since the internal 0.1 build that must have been your last experience of the browser, I suggest you give it a try.

 

Who wants social network integration in their web browser anyway? If I wanted it I'd install an extension, but oddly enough I find the actual websites enough. Or dedicated apps.

In this day and age, my car should not have got a flat tire!

Quote:

Originally Posted by RichL 

 

Chrome isn't based on Safari, it's based on WebKit. WebKit is an Apple-developed open source rendering engine used by practically every web-browser apart from IE and Firefox.

Why did Apple develop WebKit?

Was it:-

a) for a new browser called "Safari"

b) something to do, because it was fun

c) an evil way to collect more "apple tax" from iSheeps

I must say this has been an enjoying thread, a few people seem to understand the difference between user and system-land. The rest seems to be ignorant morons!

Drivers such as the graphics driver whole point in existence is to abstract the hardware through an API. Google Chrome and for that sake any other piece of software which will use the same pattern of accessing will also run into the same Apple bug. One cannot be so oblivious to believe that Chrome is the only binary artifact that is causing the bug, it is probably the most common software but by no means alone.

By no means do I blame Apple either, all software contain bugs! But the objective (as opposed to the subjective) title for this post ought to be "Google discovered Bug in Apple's kernel code causing MacBook Airs to crash". 

Google can work around the problem, we in the software business do that every day. But to work around a Driver Bug is like stepping Back to the dark ages... and is hardy the correct measure.

So now I need to find another reason why my Win 7 machine continues to freeze and blue screen, because "Windows 7 is so reliable"?!  I keep hearing people say how much more reliable Win 7 is, yet as a daily user, I find Win 7 incredibly unreliable and an operating system that requires a substantial amount of maintenance just to keep the bloody thing running.  I don't think there has been a single week which hasn't required a reboot of my Win 7 machine, including, yes, system freezes.  

Over the past year, the ONLY two issues I've had with my MBA have been (a) the occasional freezes, apparently caused by Chrome [which I've stopped using for now], and (b) some annoying Thunderbolt Display problems which now seem to be lingering in the past.  I don't know that the freeze issue that I'm having is a kernel panic, per se, because the system continues to operate with cursor movement, but I simply can't do anything (as opposed to the traditional grey-screen kernel panic which literally halts the machine with a pretty message).  

Regardless of how applications "should" behave, and errors "should" be handled by the underlying operating system, these things are complex beasts, and both Windows 7 and OS X can benefit from improvements to ALWAYS prevent applications in user-land from killing the machine.

Quote:

Originally Posted by coolfactor 

 

A lot of people are using it without even realizing it. I've spoke with a number of people for tech support that claim to be using "Internet Explorer", only to find out that they unknowingly switched to Chrome at some point without realizing it. A lot of people are drones when it comes to this stuff. And Google has been very aggressive about pushing Chrome onto people.

How does one just get switched over without them knowing lol?