Of Safari Update 2.0.1
As you're probably aware, earlier today Apple released Safari Update 2.0.1. A few notes...
- This update does not fix the Safari Image of Doom™, although I'm happy if it doesn't crash for you with however you're viewing it, but more on that in a moment.
- It's hard not to be amused that a "Safari Update" requires a restart. It's beyond silly to have to restart your computer to update your browser, which means we have to assume this is actually a global webcore/webkit update, and will affect everything depending on it -- including things like Dashboard and third party apps. I hate having to assume, which brings us to...
- You may have noticed the big box saying what's contained in this update is a little anemic. There's more info on the nutrition panel on the back of the box of gum on my desk than there is about what this update actually does, and while there's no point in beating a casual user over the head with things they won't care about, there really needs to be a link to more information.
I'm not going to start tagging The Safari Image of Doom™ onto the end of every post just yet, because it's really just been a few days since I put that post out there. Chances are this patch was already in its testing cycle -- whatever that may be -- and while there's something to be said for stopping and making sure a crashing bug that could be used in this way gets in, 10.4.3 is hopefully on the way, or more likely a security update, and they're thinking they can just roll it into that.
However, while my post just went up a few days ago, I can tell you that Apple received The Safari Image of Doom™ several months ago via the crash reporter built into OS X, and since then it's received it many times via the same system.
I know, because as part of my pet project to try to map Apple's QA from the outside I've been sitting on it for awhile, to see whether it'd get taken care of via normal channels. It's one of the reasons why I'd install and uninstall APE and such, and I've personally crashed Safari on a few systems numerous times, sending in crash reports from different user accounts and systems, often while changing the image's file name and the location of the server. Others have done the same to help me out, and I'm aware of users who sent in crash reports dealing with the same bug.
I'll let you connect the dots on that as you see fit, but the term "placebo" is coming to mind.
If it's not in 10.4.3 or something else soon, then I'll start tagging it onto posts, but will be sure to give a warning first and links to alternative browsers while apologizing to those viewing the full feed via a webkit-based RSS reader -- it's just a little early for that.
Posted by drunkenbatmanAugust 29, 2005, at 11:51 PM
Comments (32)
Posted by: Simone Manganelli at August 30, 2005 12:28 AM
Meh. I can't say I'm surprised. Apple probably gets millions of these crash reports that really help with nothing at all, because they're caused by third-party applications, or they're caused by corrupted prefs, or they're caused by a bug in an APE module, or they're caused by _____.
Basically, crash reports seem pretty useless unless someone comes in and boils it down to a test case that reproduces the problem 100% of the time or at least >50% of the time. When I have users that submit problems with regards to my software, 99% of the time I have to ask for their prefs file or for something else that causes the crash. Crash reports by themselves don't help, but in tandem with other information, they do. But with the vast majority of the "regular channels," Apple doesn't get the extra info it needs. Sending pref files with the crash reports might increase the usability of the reports, but then some users might not appreciate that, seeing as some pref files will probably contain sensitive information like account names and stuff for iChat.
Most "regular channels" are probably useless to Apple, and they really need someone dedicated enough to file a Radar bug for it to actually go and get fixed -- and even then I'm sure they have to wade through hundreds of duplicates and bugs that are really just one-time hardware failures. The signal-to-noise ratio for Radar bugs is probably much higher than the STN ratio for normal crash reports, which is why you probably haven't seen anything get done through these channels.
Of course, this doesn't propose a solution to the problem, because this is a problem. But the solution is probably enormously complex, because figuring out exactly what is wrong either requires time on the user's part, or some extremely intelligent automated troubleshooting program that's probably not built yet. So I don't really expect anything to change on this front anytime soon.
-- Simone
Posted by: Mitz Pettel at August 30, 2005 01:03 AM
That Image of Doom bug is an ImageIO bug, so it makes little sense to expect it to be fixed in a Safari/WebKit update. In fact, if you download a nightly build of the CVS version of WebKit you'll see that it "still does not fix" that bug, simply because it's not a WebKit bug.
Posted by: Romain at August 30, 2005 01:05 AM
Indeed, this update is both for Safari itself AND the Webkit. After all Safari is nothing without Webkit.
I like to have my Apps into folders like "Webdesign" or "Multimedia" within the "Applications" folder. It means that Apple's automated update process systematically looks for, say… Safari.app at the root of "Applications" where it is not and copies there the updated components, creating a non-functional App.
So, I have to momentarily put my soon to be be updated Apps in their default place in order for the process to work and thus know what is about to be updated.
If you hate having to assume what's going to be updated here is my recipe:
1) Instead of launching the process via Software Update, go download said update from Apple's support site.
2) Look at the package's structure either via right-click + "Show package content" or by using Pacifist.
3) If you are comfortable with it, launch the updater...
Posted by: joel at August 30, 2005 01:07 AM
Same here: Oh new Safari update. Great.. Starting Software Update.. hmm, reboot required?.. Let's see the description.. click.. wtf..
Posted by: Mark Yin at August 30, 2005 01:28 AM
"That Image of Doom bug is an ImageIO bug, so it makes little sense to expect it to be fixed in a Safari/WebKit update."
If Safari is crashing I expect an update to Safari to fix it. It is unclear to me if it is a bug in ImageIO or how WebKit is using it, does anyone know? They have to do something, this is a bad bug.
Posted by: Z. D. Smith at August 30, 2005 01:28 AM
'while apologizing to those viewing the full feed'
Is there really a full feed? I only see the partial one. If there is one, I'd love to know how to subscribe to it.
Posted by: at August 30, 2005 01:34 AM
Crasher Reporter is worse than useless. Don't bother to enter a detailed description, a human rarely sees them. Cozy up with a developer and get him to enter them into RADAR.
Posted by: Cap'n Hector at August 30, 2005 01:57 AM
There is a full feed…for people that donated to DB.
Posted by: Rosyna at August 30, 2005 02:40 AM
If Safari is crashing I expect an update to Safari to fix it. It is unclear to me if it is a bug in ImageIO or how WebKit is using it, does anyone know? They have to do something, this is a bad bug.
Uhm, ImageIO handles the decoding/encoding of images on Tiger. This is a good thing as in Panther, there were multiple (3+) implementations of the PNG decoder, this was bad... So if it crashes in Safari, it'll crash everywhere else as well. Just because you're only likely to see it in Safari does NOT make it a Safari bug.
Furthermore, I can't imagine each crash log is looked at. However, I can imagine the are compared with each other to try to find common crashers many people are having. Then they'll read the comments to find out how to reproduce the crash. So just because you submitted it via CrashReporter doesn't mean Apple "knows" about it. However if a thousand people have the same crash, then Apple will know about it. The best way to make apple know is via RadarWeb.
But this ImageIO bug is actually a Security bug. Even though you couldn't likely exploit it, it's a crash due to bad input data.
Posted by: 2.STL-DX.B11 at August 30, 2005 05:49 AM
Yeah - I was irritated that I needed to restart after this. I guffawed and snooted loudly at it but then had to do what Mr. G5 told me to and hit that restart button. Pah.
As a whole, I think that Apple's QC these days is in a different (for more unpleasant) place than it was about 10 years ago. Once the Intel thing happens I only see it getting worse and if they ever did decide to release OS X for all Intel based boxes (I know that that's a whole debate in itself) then the gates of hell (no pun intended) would well and truly be opened for good.
God, I'm such a cynic.
Posted by: robert at August 30, 2005 02:45 PM
"As a whole, I think that Apple's QC these days is in a different (for more unpleasant) place than it was about 10 years ago."
not to pick on the poster, but the tone of this I believe is inline with where DB has been going lately, and strikes me as a meaninngless comparision, akin to our understanding of physics as described by newton compared to einstein.
What is with this whole DB rant about apple bugs and QC? I think a good way to get page hits. Move on to a real target.
Posted by: at August 30, 2005 04:11 PM
"What is with this whole DB rant about apple bugs and QC? I think a good way to get page hits. Move on to a real target."
I doubt it, every wired Mac user already knows who he is. He's a rock star in Mac, and celebrities use their fame to focus on issues they care about. Beats MacGyver and his baby seals.
Posted by: Widget at August 30, 2005 04:45 PM
The image of doom is most likely a Safari bug, not a system "i/o" issue.
You can open IOD (image of doom) in Preview. There you will find that IOD is actually a 17 frame animated gif file (thus, in reality image(s) of doom). Safari displays the first frame then bails.
Quicktime 7 can open and play IOD correctly and doesn't crash.
Realplayer 10, which uses webkit, can open and play IOD, but jumbles the animation. However, it doesn't not crash.
Text edit (which, IIRC, uses webkit) accepts IOD via drag and drop in a new doc window, but only displays the first frame without animation. (Ordinarily, Text edit can and does animate gifs correctly). Textedit does not crash.
Mail, which also uses webkit, will accept IOD via drag and drop on a new message window. It displays the first frame only and doesn't animate (I'm not sure if mail normally animates gifs files; in the past my experience has been that it does not). However, Mail does not crash.
By the way, Safari updates generally deliver updated webkit/webcore files, which permeate the system, i.e., other apps such as mail, widgets, textedit, etc., rely on webkit for certain functions.
Posted by: drunkenbatman at August 30, 2005 05:34 PM
I doubt it, every wired Mac user already knows who he is. He's a rock star in Mac
Oh Lord, lets nip that in the bud right now. I do appreciate the sentiment, but please don't go around saying that sort of thing.
Posted by: Kurt Moore at August 30, 2005 06:46 PM
I made a little bash script to keep an eye on what gets changed when I install software, ran it before I ran the software update for Safari.
While the script simply touches a file for a time and date stamp and then finds all newer files, its been pretty accurate. I did not have anything else going on at the time, other than a IM conversation, which I stripped out of the log.
A ton was changed, so its more than just a Safari update, at least, there were a lot more items updated.
Posted by: Daniel Steinsland at August 30, 2005 07:03 PM
I believe you are wrong. At least on my system it DOES in fact fix the image of doom.
Posted by: drunkenbatman at August 30, 2005 07:06 PM
I believe you are wrong. At least on my system it DOES in fact fix the image of doom.
And I believe anyone ruling what happens with their software is the standard for everyone else simply hasn't been reading the site long enough.
Posted by: Daniel Steinsland at August 30, 2005 07:49 PM
I've been reading your site for a long time, and I agree with you. However, it's about semantics. You claim in your post that the update does not fix the IOD period. This is misleading. I understand that it might continue to be a problem for some people, but for others it's now a non-issue.
I love your blog and read it religiously, believe it or not... :-)
Posted by: gevin at August 30, 2005 08:08 PM
hey, safari ver.1 user like me also got new update! :)
Posted by: Hao at August 30, 2005 08:16 PM
I was all upset when I tried the IOD and Safari didn't crash immediately. Then I hit reload and the world was back to normal. I'm still tempted to send that image in an e-mail to my house's spam list and hear the cries of horror from the other mac users... :)
Posted by: DaveB at August 30, 2005 08:32 PM
Don't know about your IOD crash but the (1.3.1) update has fixed a (javascript?) crash I was seeing @ hse.gov.uk.
Posted by: drunkenbatman at August 30, 2005 10:17 PM
I've been reading your site for a long time, and I agree with you. However, it's about semantics. You claim in your post that the update does not fix the IOD period. This is misleading. I understand that it might continue to be a problem for some people, but for others it's now a non-issue.
*grins* I get what you're saying, but this bug wasn't 100% reproducible by everyone before the update, so I'm not surprised when someone isn't able to reproduce it after the update while others are.
Posted by: Crazy Ernie at August 31, 2005 03:02 AM
Oh Lord, lets nip that in the bud right now. I do appreciate the sentiment, but please don't go around saying that sort of thing.Not a rock star; so what, then, you'd rather be a jazz icon? Superhero? Candidate for the vice-presidency? Come on, you know we revere you now.
(And Apple QA is a way better topic to spend your celebrity on than baby seals.)
Posted by: DaveB at August 31, 2005 04:50 AM
Have they improved Safari's caching with this update?
(1.3.1)
I have dial-up internet access and I am noticing a marked increase in page load speed when opening multiple tabs from the same website.
Posted by: Uli Kusterer at August 31, 2005 06:11 AM
Just wanted to chime in saying I also had no problems with you image of doom (at least pre-2.0.1 -- haven't tested with that yet). Even a re-load didn't shoot down the browser. I think you may want to get a (free) ADC Online account so you can use Apple's Bug Reporter to report this bug. Then you can attach an Apple System Profiler report of the systems it crashes on. That tremendously seems to help Apple engineers figuring out crashes.
Posted by: Sören Kuklau at August 31, 2005 07:21 AM
I made a little bash script to keep an eye on what gets changed when I install software, ran it before I ran the software update for Safari.
Um, you *do* realize that you can just run 'lsbom' on the .bom file inside the installation package and get a clean version of that exact list?
Posted by: J.D. at August 31, 2005 08:29 AM
"I was all upset when I tried the IOD and Safari didn't crash immediately. Then I hit reload and the world was back to normal. I'm still tempted to send that image in an e-mail to my house's spam list and hear the cries of horror from the other mac users... :)"
Same thing here. On first load nothing happened. Reloaded it and WHAM! Down goes Fraizer!
Posted by: Troy at August 31, 2005 08:30 AM
I don't get crashes with the image, but I still on 10.3.9 and just updated Safari via Software Update?
Hey my comment post failed if I used my blog URL (free one owned by Google) with: Your comment could not be submitted due to questionable content
Posted by: drunkenbatman at August 31, 2005 03:00 PM
I think you may want to get a (free) ADC Online account so you can use Apple's Bug Reporter to report this bug. Then you can attach an Apple System Profiler report of the systems it crashes on. That tremendously seems to help Apple engineers figuring out crashes.
If they removed the requirement of signing an NDA just to report a bug into their system or download updates to GCC and other tools, I would.
Posted by: Daniel Brauer at August 31, 2005 11:16 PM
Maybe your machine is unstable because of your interface modifications. That screenshot was not taking in a vanilla install of Tiger.
Posted by: Chris at September 1, 2005 08:23 AM
Crash! Boom! Bang! (Roxette and Safari 2.0.1).
No crashy for me.
Found out about the update from journals.ars, installed it, had the same amusing thought about the requirement to reboot, and headed straight here.
Sometimes these bugs can get missed I think if they can't replicate. At a former workplace, I rolled out a lab of lamp iMacs at about the time that 10.3 was close to release. We had an LDAP directory server running on Linux providing authentication, all was nice, except that for some reason, the Macs would allow you to log in without a password.
Type in a valid username, no password, and you get access to the users account and files. We contacted Apple, they sent out 3 engineers, they took network dumps, the configuration of our LDAP server, witnessed the problem in action, and came back with "Can't replicate problem internally."
Suggested upgrading our version of LDAP, or waiting for 10.3 to come out.