Oh i know a hardware problem! I have an iSub and the @!#$% thing never works right anymore (stupid USB-connecting iSub)
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Mac OS X 10.4.3 8F28 Seeded
- Thread starter MacRumors
- Start date
- Sort by reaction score
I'd got further and get rid of that "almost".SiliconAddict said:
Given how much modern operating systems depend on things like hardware abstraction layers and virtual machines and microkernels, it should be possible for an OS to trap a crashed driver and remove or restart it.
Crashes in the face of hardware failures, of course, are another thing. Avoiding system crashes in that situation would require redundancy. Some of this is in Mac OS right now (like RAID support), but other aspects (like disabling failed processors or memory modules without crashing the system) would probably require different core-logic chips.
That's the idea of a microkernel, right? Unfortunately in practice, most microkernels are crap for performance. OSX uses a hybrid micro/modular kernel and it is still said that it takes a huge performance hit for that very reason. The HAL helps to keep apps from making a wrong call directly to hardware (if the app developer decided to code to the HAL, and not some unpublished kernel API), but really has nothing to do with a badly written driver.shamino said:
I don't know if we can have properly jailed drivers and still expect a lean, quick OS. At least, I don't think we can expect this yet...
Jim
Considering how much CPU power everybody wastes on eye candy, I would find it hard to believe that this is at all a factor.jim. said:
I'd much rather give up all the gratuitous GUI effects and have a crash-proof kernel. I'm sure apps would end up at about the same speed they run at now.
Moore's law. CPUs are always getting faster. If performance takes too much of a hit now, it will be just fine in a few months.jim. said:
Microsoft uses this logic to justify the ever-increasing amount of bloatware in their OS and applications. Ditto for a lot of the popular games.
So why can't this argument be used to make a kernel change that will make systems virtually crash-proof?
I can only assume that people just don't care all that much for stability, no matter what they tell themselves.
One would think, but the microkernel design has been plagued by this performance problem on desktops for nearly 20 years. I think it is easy to have a redundant, and nearly crash proof kernel if your specifications for use are clearly outlined (Embedded systems, for example). Part of the problem with desktop use is, for every advancement in hardware, there is a corresponding change in paradigm as to what should be done on a computer. There is no one set use, and these new programs and hardware need to perform hacks directly on the kernel, or in hardware, sometimes to perform desired functions.shamino said:
It is very rare that a mature kernel just core dumps because of itself. It is typically because of a bad call directly to hardware from something, or of bad hardware. The way to eliminate these would be true microkernel design (and program watchdog) with all of the inherent problems in message passing, or jailing all programs and drivers while trying to develop a secure message passing mechanism between jails that can't be exploited, once again taking a huge performance hit.
Unfortunately as long as the potential of a desktop computer is changing, Moore's law just allows us to stay afloat with hardware requirements. Lean GUI's would help some, but probably not a huge amount, because you're talking about a slowing down in hardware communication for low level functions, which are not affected by what is going on above.
Jim
shamino said:
I agree with you in principle, but the post-2003 performance world looks rather different from the pre-2003 one. We're still getting lots of transistors (so technically Moore's law works), but they're no longer getting much faster and cooler with each shrink. The G5's scaling problems and Intel's abysmal revision to the Pentium-4 demonstrate this nicely. Multicore will help a lot, if programmers can figure out how to multithread everything enough.
tiger is good/bad
I have had many problems with this OS, the experience hasnt been better than panther which is rock solid up to .7 then they screwed up bad on .8 and saved it with .9
tiger has been way too buggy even though it IS a BIG STEP forward.
I'm not sorry I installed the Dec. beta... on a separate drive of course...
but yeah, keep the updates coming and give me back my Finder/mail/safari/firefox stability
I have had many problems with this OS, the experience hasnt been better than panther which is rock solid up to .7 then they screwed up bad on .8 and saved it with .9
tiger has been way too buggy even though it IS a BIG STEP forward.
I'm not sorry I installed the Dec. beta... on a separate drive of course...
but yeah, keep the updates coming and give me back my Finder/mail/safari/firefox stability
I have the feeling that rumors of new updates/fixes tend to draw out all of those who are having problems with their install, or at least feel passionate about the problems they've been having. i've never experienced any bugginess... then again, i probably just jinxed myself right there.
take your blind folds off.
i truely think that how tiger is right now is up there with windows xp. i have atleast 1 problem a day that makes me have to reboot. my friend who just bought a 17" g5 imac cant even use tiger because the whole system will freeze after a few minutes of use. i have more friends that have tried tiger and went back to panther because of all the problems. honestly, i thought i moved to apple to get away from this. without stability os x is just a pretty looking operating system.
i truely think that how tiger is right now is up there with windows xp. i have atleast 1 problem a day that makes me have to reboot. my friend who just bought a 17" g5 imac cant even use tiger because the whole system will freeze after a few minutes of use. i have more friends that have tried tiger and went back to panther because of all the problems. honestly, i thought i moved to apple to get away from this. without stability os x is just a pretty looking operating system.
SiliconAddict said:
Thanks to Apple then...that's why I've had NONE so far with my iMac G5...the fact that some people have problems (with whatever config and 3rd party add-ons they might have) does NOT mean the System is flawed...
Apart from a few cosmetic bugs, I have had a GREAT experience since 10.4.0...and the developers out there are smart enough to iron out bugs, no need to spread OS X out to the whole world...that's why Linux is such a ridiculous mess.
yeah ive tried that a few times but ive come to the conclusion that its not something i can fix, its bugs. for example lets say im dragging something then open expose by mistake, everything will freeze up. sometimes if im using 2 applications at once itll freeze up.
if it was just me having the problems i would understand, but everyone i know who has tiger is having problems with it....well i think theres 1 that says he doesnt have problems, but he doesnt really do much with it excpet go online and listen to music. i do have faith that it will get better but i dont think they should be releasing anything that buggy. if they start doing that they are no better than microsoft(i hate to say it). i always thought apple = quality.
if it was just me having the problems i would understand, but everyone i know who has tiger is having problems with it....well i think theres 1 that says he doesnt have problems, but he doesnt really do much with it excpet go online and listen to music. i do have faith that it will get better but i dont think they should be releasing anything that buggy. if they start doing that they are no better than microsoft(i hate to say it). i always thought apple = quality.
10.4.2 GOOD
I did a "clean" install of Tiger and have not had any problems at all other than an occasional application crapping out. I use Maya7 Unlimited, Motion, modo, Zbrush, Vue5 Infinite, PhotoshopCS, FCE on a daily basis and this machine is very solid.
I don't understand what these "microkernal" problems are and whatnot you tech experts are talking about. All I know is I push my Dual2.0 hard with expensive software and the OS makes it all work for me!
I did a "clean" install of Tiger and have not had any problems at all other than an occasional application crapping out. I use Maya7 Unlimited, Motion, modo, Zbrush, Vue5 Infinite, PhotoshopCS, FCE on a daily basis and this machine is very solid.
I don't understand what these "microkernal" problems are and whatnot you tech experts are talking about. All I know is I push my Dual2.0 hard with expensive software and the OS makes it all work for me!
A microkernel is not a problem, but a kind of system design.bernardb said:
With your classic "monolothic" kernel, the core of the operating system runs (more or less) as a single large program. Different pieces of this program manage different parts of your hardware (like memory, disk drives, etc.) If it crashes, the entire system crashes.
In a microkernel, the operating system's core runs as a collection of separate programs. They are isolated from each other just like applications are. They talk to each other through well-defined messaging systems, just like applications. If one of these programs crashes, the rest of the system won't crash, and the failed program can be auto-restarted.
The upside to a microkernel is that it is much more flexible and stable. The downside is that it is less efficient - all that messaging consumes CPU time that would otherwise be spent running application code.
Most modern operating systems (including OS X) are sort of a hybrid, with some stuff operating in a monolithic kernel, but with some system facilities operating in separate processes. Determining what runs where, in order to balance system stability against performance, is one of the key jobs of an OS designer.
Core Apps still broken in 10.4.2
I did an upgrade of a fully functional 10.3.9 to 10.4 then immediately applied the combo patch to bring me up to 10.4.2. That's when the problems started. From what I can see there are major problems with the underlying web kit that makes Safari work.
As soon as I applied the 10.4.2 Combo patch and tried to run Software Update it would not work. I could see the available updates, but when I tried to install them I would receive some error. I tried everything: repaired permissions, contacted AppleCare--No luck. After a few days Software Update started working again. Strange.
On the same day of my upgrade Safari could access any site on the Internet except Apple.com! What gives?! I worked with AppleCare to reset Safari, delete cache files, delete preference file, and even reinstall Safari no luck. Again, Safari strangely started working with Apple's site a few days later.
Then Safari started acting strange. After what seems to be about 10 mins of browsing I start receiveing these messages:
Safari can't open page. the error was: "lost network connection" (NSURLErrorDomain: -1005)
-or-
Safari can't connect to the server. Safari can't open the page "..." because it could not connect to the server "..."
I have applied Safari 2.0.1 patch and this has not helped the problem. Running all other current patches/updates from Apple.
I am >>TIRED<< of Safari refusing to work. Apple please fix this problem!
MacsRgr8 said:
I did an upgrade of a fully functional 10.3.9 to 10.4 then immediately applied the combo patch to bring me up to 10.4.2. That's when the problems started. From what I can see there are major problems with the underlying web kit that makes Safari work.
As soon as I applied the 10.4.2 Combo patch and tried to run Software Update it would not work. I could see the available updates, but when I tried to install them I would receive some error. I tried everything: repaired permissions, contacted AppleCare--No luck. After a few days Software Update started working again. Strange.
On the same day of my upgrade Safari could access any site on the Internet except Apple.com! What gives?! I worked with AppleCare to reset Safari, delete cache files, delete preference file, and even reinstall Safari no luck. Again, Safari strangely started working with Apple's site a few days later.
Then Safari started acting strange. After what seems to be about 10 mins of browsing I start receiveing these messages:
Safari can't open page. the error was: "lost network connection" (NSURLErrorDomain: -1005)
-or-
Safari can't connect to the server. Safari can't open the page "..." because it could not connect to the server "..."
I have applied Safari 2.0.1 patch and this has not helped the problem. Running all other current patches/updates from Apple.
I am >>TIRED<< of Safari refusing to work. Apple please fix this problem!
You Safari problem is more likely related to a known error in Lookupd. Seems that it is some type of network component and when it hangs, it slowly starts to hose the entire system, thus requiring pressing the ol' reset button to fix. Supposedly, if you can disconnect, then reconnect to the network, this will also restart the Lookupd service.
From MacNN Sept 27, 2005:
Unlockupd 1.0 (free) works around a bug in lookupd, a system service which is required for proper operation of Mac OS X. When lookupd fails, the system quickly becomes unusable. Unlockupd periodically checks lookupd's status and forces it to restart should it fail.
This had BETTER be fixed with 10.4.3 -- it's embarrising having to restart my Mac with a reset button....This Unlockd 1.0 seems to have fixed (work around) my problem for now. No resetarts required in last 3 days since installing it.
From MacNN Sept 27, 2005:
Unlockupd 1.0 (free) works around a bug in lookupd, a system service which is required for proper operation of Mac OS X. When lookupd fails, the system quickly becomes unusable. Unlockupd periodically checks lookupd's status and forces it to restart should it fail.
This had BETTER be fixed with 10.4.3 -- it's embarrising having to restart my Mac with a reset button....This Unlockd 1.0 seems to have fixed (work around) my problem for now. No resetarts required in last 3 days since installing it.
wdlove said:
I usually use the following blog for that type of info:
http://haligon.blogspot.com/
AppleMatt
