Mac OS X 10.4.7 Seeded

[MacRumors]

Apple has seeded Mac OS X 10.4.7 to Apple developers earlier this week. The current build is reported to be 8J111 for PowerPC and 8J2111 for Intel Macs.

Areas of testing are listed as Mail, Safari, Aperture, Syncing and iChat. Known issues include inaccurate Calculator results and Mail signatures note working reliably.

[Roller]

Lots of updates lately - that's generally good, although I think that 10.4.6 made my system a bit less stable.

[kwajo.com]

i've been very happy with all the versions of Tiger on all 5 of my machines, and 10.4.6 seems to be the most reliable yet. Can't wait for this next update, I hope it does something about energy saving settings. A couple of my systems, including my G5, have trouble going to sleep when they're supposed to

[thomasp]

That "inaccurate calculator" bit intrigues me?

*Enters '1 + 1' into calculator*

*Hits 'Equals'*

*Calculator returns result of '3'*

[bowzer]

Originally Posted by Macrumors

Known issues include inaccurate Calculator results

You'd think calculator would be a pretty easy app to get right...

[sam10685]

Originally Posted by bowzer

You'd think calculator would be a pretty easy app to get right...

i agree... a computer IS a calculator so that app should be pretty easy to nail.

[miketcool]

Calculator?

Makes those new Mac commercials kinda silly now. iLife, cool. At least the calculator app on windows works correctly!

[m-dogg]

Originally Posted by kwajo.com

i've been very happy with all the versions of Tiger on all 5 of my machines, and 10.4.6 seems to be the most reliable yet. Can't wait for this next update, I hope it does something about energy saving settings. A couple of my systems, including my G5, have trouble going to sleep when they're supposed to

Y'know, my iMac G5 has been having some sleep issues lately as well. Never associated it with that update, but it probably was around that same time...

[Macnoviz]

Originally Posted by bowzer

You'd think calculator would be a pretty easy app to get right...

I read about such things once, but don't worry we're talking about a difference of 0,00000000000000000000001 or so (not really sure about the amount of zeros)

These differences are really no big deal for almost any user, but they can make a difference in science (especially space and such), unfortunatly, one of the Powermacs (and other macs) selling points.

[ToastyX]

I'm sick of how buggy Mac OS X is, but nobody seems to notice because they're too busy claiming how "superior" everything is.

I complained about the calculator back in November: http://forums.macrumors.com/showthread.php?t=160617

They're just now getting to it?

There are also tons of other bugs that aren't getting fixed, but nobody seems to notice or care.

[joshysquashy]

do you think newer phones will be supported in isync? is this what syncing refers to?

[Mr_Brightside_@]

How about the fact that the controller bar in DVD player disappears when you go into expose? that's been there since the start, and is quite annoying. as for calculator, it could have something to do with order of operations.

[LastLine]

Originally Posted by kwajo.com

i've been very happy with all the versions of Tiger on all 5 of my machines, and 10.4.6 seems to be the most reliable yet. Can't wait for this next update, I hope it does something about energy saving settings. A couple of my systems, including my G5, have trouble going to sleep when they're supposed to

Y'know, my iMac started doing that since the 10.4.6 update too - I found resetting the PRAM actually helped sort it (I think the phrasing's right there) and it's not done it since (touch wood). Might be coincidence, but worth a shot in my opinion.

[rye9]

Originally Posted by kwajo.com

I hope it does something about energy saving settings. A couple of my systems, including my G5, have trouble going to sleep when they're supposed to

Yupp, same here, I emailed Apple last month about this too, hoping the next OS update would fix sleeping problems. I just ended up resetting my PMU.. it worked and now it's fine though.

[ChrisA]

Originally Posted by bowzer

You'd think calculator would be a pretty easy app to get right...

Really? How would YOU propose the TEST it?

Let's see now I could add "1" to every number betwenn 0 and 10 billion and when I'm done I'll try adding 2, then 3 and so on untill I've done a billion plus a billion. Then maybe I try negative numbers then move on to division and square roots. If I do 10,000 tests per second could I actually finish before the sun runs out of hydrogen fuel to burn. Gosh no.

So it seems testing is not the way to go here.

How about spot checking? Here is a good example -- I have a new theorem: "60 is divisable by all integers?" Let's try: 1 goes into 60?, Yes., 2, yes, 3, yes, 4, yes, 5, yes, 6, yes. this is to slow, 10, yes, 20, yes , 30 yes. end of proof: "60 is divisable by all integers.

OK so much for spot checking.....

So if it's "easy" tell us how you whould do it and if your method would find the problem where the square root of pi/0.334 is wrong in the 6th decimal place.

Actually checking a calculator is used as a classic example of a hard problem. How to find that one in a trillion error in the 6th decimal place and you can't exaustive test and you can't spot check. You are prettymuch stuck with a functional analysis of the design and what they call "white box" testing.

[joshysquashy]

Originally Posted by ChrisA

Really? How would YOU propose the TEST it?

Let's see now I could add "1" to every number betwenn 0 and 10 billion and when I'm done I'll try adding 2, then 3 and so on untill I've done a billion plus a billion. Then maybe I try negative numbers then move on to division and square roots. If I do 10,000 tests per second could I actually finish before the sun runs out of hydrogen fuel to burn. Gosh no.

So it seems testing is not the way to go here.

How about spot checking? Here is a good example -- I have a new theorem: "60 is divisable by all integers?" Let's try: 1 goes into 60?, Yes., 2, yes, 3, yes, 4, yes, 5, yes, 6, yes. this is to slow, 10, yes, 20, yes , 30 yes. end of proof: "60 is divisable by all integers.

OK so much for spot checking.....

So if it's "easy" tell us how you whould do it and if your method would find the problem where the square root of pi/0.334 is wrong in the 6th decimal place.

Actually checking a calculator is used as a classic example of a hard problem. How to find that one in a trillion error in the 6th decimal place and you can't exaustive test and you can't spot check. You are prettymuch stuck with a functional analysis of the design and what they call "white box" testing.

I think you misunderstood him: computers are calculators so a calculator function is the easiest app to build, compared to say, image editing. Image editing uses mathematical calculations to create and manipulate images, much more complex.

any programmer (I am not one) knows that a good place to start programming is to create a simple calculator. the difference here is that apple are looking at inaccuracies in incredibly complex calculations which take the computers to their limits.

[plinden]

Originally Posted by bowzer

You'd think calculator would be a pretty easy app to get right...

You forget that while people use decimal, computers use binary. All user-entered numbers have to be converted to binary for the computer to actually perform the calcution. Integers are easy (8000 = 1111101000000) but floating points are hard. How would you go about representing 3.142 in binary? And that results in rounding errors that can be very noticeable if not taken into account.

[jamesmcd]

Well both my calculators work perfectly so I don't know what ya guys are on about.

[Doctor Q]

Some of you may remember that Calculator.app Security Update 2006-007 was issued exactly 46 days ago. :wink:

[Marlon_JBT]

Originally Posted by jamesmcd

Well both my calculators work perfectly so I don't know what ya guys are on about.

Mine does too?

What's going on here?

[SiliconAddict]

Originally Posted by ToastyX

I'm sick of how buggy Mac OS X is, but nobody seems to notice because they're too busy claiming how "superior" everything is.

I complained about the calculator back in November: http://forums.macrumors.com/showthread.php?t=160617

They're just now getting to it?

There are also tons of other bugs that aren't getting fixed, but nobody seems to notice or care.

Every OS sucks. Windows, Mac, Linux, etc. The thing is you still can have a favorite even though it sucks. After a few months with OS X now I still like XP better then OS X. But I like the apps on OS X more. There is a trade-off on everything when it comes to computers. Apple has some strengths, Windows has others, and Linux has others as well.

[maxvamp]

Originally Posted by ChrisA

Really? How would YOU propose the TEST it?

If it were me, I would likely write an automation program that would tie into the methods embedded in the Calculator App, and submit values to the methods, and compare the output with values I believe they should be. I would also do some spot checking to verify manually that they app behaves correctly.

Yes, some of this is blackbox testing, and any errors found through the automation app would be suspect until the test app has the provability of time, but even then, errors would draw human attention.

Unit tests may help, but I am not an ultra huge fan of them when compared with other forms of testing.

Max.

[ChrisA]

Originally Posted by maxvamp

If it were me, I would likely write an automation program that would tie into the methods embedded in the Calculator App, and submit values to the methods, and compare the output

That's called "spot checking".
It would not have found these obscure calculator errors nor would that have found the now infamous Pentium floating point bug. The Pentium bug was caused by some select few floating point values that just happened to have a large number of ones in a row in their binary representation and even then you had to have the "right" sequence of events to trigger it. Automated testing would not have caught that in a reasonable amount of time.

I think there are only two options
(1) Write the software in such a way that you can _prove_ it's correctness. This is very hard and requires PhD level computer scientests on staff or (2) test as best you can and live with the fact that there are likely some bugs not yet discovered. Going one step futher you can use some metrics to estimate the number of undiscovered bugs and release the software when that number is below some threshold.

[maxvamp]

Originally Posted by ChrisA

That's called "spot checking".
It would not have found these obscure calculator errors nor would that have found the now infamous Pentium floating point bug. The Pentium bug was caused by some select few floating point values that just happened to have a large number of ones in a row in their binary representation and even then you had to have the "right" sequence of events to trigger it. Automated testing would not have caught that in a reasonable amount of time.

I think there are only two options
(1) Write the software in such a way that you can _prove_ it's correctness. This is very hard and requires PhD level computer scientests on staff or (2) test as best you can and live with the fact that there are likely some bugs not yet discovered. Going one step futher you can use some metrics to estimate the number of undiscovered bugs and release the software when that number is below some threshold.

Actually, it's called Functional and / or System tests.

You would need to ( and could ) write your software to check in such a way that it could have even detected the Pentium bug ( think networkability and long run automation ) .

As one who has made such tests, I can tell you that it would not be that hard to do.

Now, due to the price charged for this app, is it worth as much effort as, say, Testing the file system????

There are acceptable levels of ship-able bugs.

No software can ever be made perfect...AND ship.

Max.

[cc bcc]

Try this in Calculator:

654,654.321 - 987.312

You'll get 653,667.00899999996 instead of the correct 653,667.009

(weird, copy pasting the result out of Calculator gave me the correct number)

It's a side effect of the binaire nature of computers.