Print

[(S)ATAman]

Only a psycho can do that!

You finally get back to a long-abandoned project and you finally can do with it what you want, free from (other) managers.
You are wrong!

You re-name that project from BluBlaBmuxTex/12DF5/67_23_67_BV_Y-4ABCD (try to pronounce it!!) into "SImagix_II"
(Hello Asterix with Magician and SIlicon IMAGe; "II" means a second-generation SATA controller)

You debug it (an of course ready to move down to "9" and up to "Catalina").
One of the critical phases is to build under 10.13.6 (just because, this is long to explain).

You kill one old bug after an other, it's so much easier now, almost fun.
Than you notice, the XCode under 10.13.6 complains about something in a warning.
After a while you realize, this is a very-very-VERY important warning rejecting the "SImagix_II"

Because... because there is a "_" character in the name of the bundle and some managers decided, it is not permitted.

You are back to change all names in all related build scripts and settings and CodeWarriors and who-knows-what.
Busy with VERY-VERY important things.

Why? Because you know the psychopaths. Few more Californian landscapes or wild cats later they will make sure EVERYTHING in your driver building will fail with a fatal error, not merely with a warning (isn't it a fun?) just because of "_" character in the bundle. So you fix it now (spending useless time) and in the upcoming "9" project, too... to save more time with frustration later.

Who the hell employs these parasite bozo-s at the "mothership"? Duh.

[IIO]

lol, but see, it could be worse: in windows OS you may not even have a "+" sign in a filename.

and there are more people hit by that than a few dozen firmware developers and ROM hackers.

Who the hell employs these parasite bozo-s at the "mothership"?

the super-psycho. there is one in every CEO.

[(S)ATAman]

the super-psycho. there is one in every CEO.

Apparently, it HELPS a company a lot if the CEO is a psycho. And even more if (s)he is a super-psycho.

Very-very bad news for me (hopefully!), my current employer... and a few ones of the past.

But on the other hand having psycho on the top is hurting your everyday life even more.

So i guess, better to survive somehow without the psychos at the tom. Even if it once in a while may hurt your finances.


Apparently, the worst is if you are ADD, your wife is ADD and your employer is ADD.
Been there, done that and still kicking. There is only one thing worse than that: your kids are ADD and your close friend, the co-founder of your company is ADD. That sounds awful and I have an eerie feeling...

[GaryN]

Who the hell employs these parasite bozo-s at the "mothership"? Duh.

It sounds to me like you're talking about THIS guy.

[IIO]

around y2k when many new employees started their career in cupertino i have heard more than one story about people who were seriosly asking themselves whether the company has already been taken over by scientology or not.

while two key cards, cameras in every room, a bullet-proof NDA and so on can be considered "normal" for an OS- and hardware developer, some other things which happened there are said to be seriously wrong at all levels.^^

...

then OSX 10.1 came finally out and it had a serious security vulnerabilty where you logged out, and then when you recalled a text document from the "last used items" finder menu you were logged in again.

thats when they reduce their stakeholders to the shareholders only. the customers and the workers are not in their focus at all.

[(S)ATAman]

around y2k when many new employees started their career in cupertino i have heard more than one story about people who were seriosly asking themselves whether the company has already been taken over by scientology or not.

Were you in Y2K time around Cupertino?
I was, even more, I used to work at a certain known company with HQ in Cupertino.

Before our entire department was Amelio-rated to make a certain gentlemen called "Steve" happy.
Who just came aboard.

My project was quite sucky to tell the truth: I had to restore a code which was broken because they lost a good part of the sources.
Ultimately I did succeed (almost) - but not before finding the "mother of all bugs" in the than-popular OS.

It was on a source codel-level two-machine debugger project by the way.

The bug was about context switching and was NEVER fixed, our entire department was fired before they would fix and than came the "wild cats" followed by "Californian landmarks".

Just FYI... if anyone feels that they have anything to do with that particular OS - there is a ticking bomb inside, ready to explode.

Normally it happens when you have two computers and you debug one from the other remotely.
That fun may even come this year, who knows.

[IIO]

i am more worried about the architecture timebombs.

would you buy a 60,000 dollar workstation for private use when the transition to ARM lurks around the corner?

[(S)ATAman]

i am more worried about the architecture timebombs.

would you buy a 60,000 dollar workstation for private use when the transition to ARM lurks around the corner?

I do not know how far around the corner it is. Cannot comment simply because I do not know to what exactly the ARM vs. Xeon vs. Ryzen race boils down.

One is sure - and this is thanks to my recent presence here and to my presence with my current employer.

While with the latest I learned that the "X" interrupt management is still way behind of Windows, Linux, Free(Net)BSD.
It sux horribly and blows.
While here I was "forced" to look at how things did look around G4 and G5. All I can say, Larry Barras (the only guy who made "9" AIM and later was hired by Apple) did see the problems very exactly back in 2003-4-5.

If you check out his (100% it's his!) code for Apple K2 SATA (same as Broadcom / ServerWorks "Frodo") than you see, Larry wrote own interrupt controller within the K2Root. It is in AppleK2SATARoot.cp file, google duckduckgo is your friend.

Look at the AppleK2SATAIC implementation!

Godfrey and Larry may not like each other but neither is stupid to say the least.
There was a mighty reason Larry made the K2Root for G5 the way he did: because he knew, interrupts do suck.

Today almost nothing did change!

They tried to fix these goddam interrupts in Yosemite. Did not work out.
Than they tried to fix them in ElCrapitan El Capitan. That did not work out either.
Than they fixed to some degree in Sierra, that at least put a very thick paint over the old bug implementation failure and it is a bit less obvious. But it's still there (as far as Mojave). I did not look into Калинка, калинка, калинка моя, https://en.wikipedia.org/wiki/Kalinka_(song) Catalina yet, but do not expect any improvement.

No developer information regarding that is updated after ca. 2011.

This is a very serious problem (called MSI-X interrupts) where any modern OS completely obliterates "X", at least on the paper.
One can write (and we see, Larry Barras prepared for that) an own MSI-X interrupt handler.

In 2020 we are still behind Larry's code. Apple's latest NVMe driver knows about MSI-X, so do I. As far as I see, I am the only person outside of Apple who has the MSI-X handler. Great... but all MSI-X handlers (either mine or of Apple) deal with the first interrupt vector only!

Which sux and blows.

Where I expect it really suck and blow would be a gazillion-core processor implementation: no matter, how many cores you have, all your driver code depends on a single interrupt vector... that means, on one single core! The IOP-s should suffer.

But no, instead all kind of mind-blowing coolness is implemented in all that... "Калинка". But as I guess, still no proper MSI-X implementation.

The best thing is that Apple wants people be less involved with writing drivers "because it's difficult" than fixing their debugging tools. Windows developers complain about Microsoft debugger called "Windbag" "Windbug". That is infinitely better to work with than with the LLDB. Shame on Apple.

[(S)ATAman]

BTW: "Frodo" was one of the best designs of it's time. It is a SATA-I (150 MByte/Sec) PCI-X chip, tolerant to both voltages.
The card was made that it fits all PCI-X slots, regardless of voltage.

The code is now in Linux, FreeBSD and as AppleK2 also in Darwin, with open source.

The chip has two faces: a traditional SATA-ATA Bus Mastering Model and an advanced "QDMA" Interface, working with so-called "Command Descriptor Ring"-s, that's why the code-name of the chip was "Frodo".


As for ARM - everything is possible, of course. The company is not driven by engineering, it is driven by marketing.
Looking at their user interface "design" once in a while  I wonder: in their everyday life do some managers use macOS there at all or not?

[(S)ATAman]

"In 2020 we are still behind Larry's code. Apple's latest NVMe driver knows about MSI-X, so do I. As far as I see, I am the only person outside of Apple who has the MSI-X handler. Great... but all MSI-X handlers (either mine or of Apple) deal with the first interrupt vector only!"

How it happened? The (obsolete) NVMe controller of Silicon Motion International (SM2260) does not expose MSI-s (Message Signaled Interrupt)-s. Instead, they provide a gazillion (32 if I recall) MSI-X interrupt vectors, without any MSI.

The NVMe driver from Apple had to deal with it and there is definitely an MSI-X handle there since Sierra.

I did try the same, discovered that SM2260-based NVMe drives do show up improperly in anything earlier than Yosemite.
With Yosemite and ElCapitan I can find the MSI-X (following Apple's NVMe code), but it won't work.

I would lose interrupts probably because the system can deal with the first vector only and the rest of it is not prepared.
But in Sierra - High Sierra - Mojave - Catalina that works, but only with the first vector. All others are ignored by the system.

So there is no point of using MSI-X unless your hardware has no MSI, only MSI-X and you HAVE to provide a Thunderbolt product as these won't pass the certification unless they implement MSI.

Than looking at Larry's 15 year old code you wonder, what went wrong at Apple.

[IIO]

sounds a bit like the sister story to the hardware bugs video recently posted. figures, since it is the same company.

it is probably based on a fundamental misunderstanding on how to solve problems or how to design software in general (first care for the most relevant parts of the base system and make it perfect, then add higher layers... every schoolkid know that... but apple, microsoft, siemens regulary fail in it because [insert random rant about the economic system here])

but it is irrelevant to argue or request things when a management decides things based on their personal interest.

can you say "native instruments"?

[(S)ATAman]

but apple, microsoft, siemens

There was a joke in DDR about Robotron but than after Unification it turned out, the very same joke is all over the "West" (western part of the "new", larger country) about Siemens. The joke does not sound good in English, so here is the original, with my bad grammar.


Ein Baby wurde morgens neben Eingang von Robotron gefunden.
Man hat das Baby sofort an Vater Staat weitergeleitet mit der Begründung, das es nicht Robotron gehört:

"Bei uns hat man nie 'was mit Lust und Liebe gemacht"
"Bei uns wurde kein Projekt in 9 Monaten erfolgreich beendet"
"Bei und hat man nie 'was gemacht, was auch Hände und Füße hat"

[IIO]

some of the GDR products like kitchen aids, factory production machines, furniture, were bestsellers here in the west and also in russia and in the USA, and today you would be lucky to be able to buy products of the same quality and reliability.

most owners in the US thought their kitchen utilities would come from western germany.

at the same time many citizen in the GDR didnt know about the good quality products either, because the were only for export.

today everything is assembled in china anyway, no matter what the "made in" label says, and usually breaks after 3 years.

https://www.ebay-kleinanzeigen.de/s-anzeige/kuechenmaschine-komet-ddr-komplett-mit-ersatzmaschine/1296240924-86-13826?utm_source=focus.de&utm_medium=cpc&utm_campaign=coop-focus.de_goods&utm_content=20181000001

look at these plastic containers. they are original parts, from 1970, and have been in use for 50 years.

[(S)ATAman]

some of the GDR products like kitchen aids, factory production machines, furniture, were bestsellers here in the west and also in russia and in the USA, and today you would be lucky to be able to buy products of the same quality and reliability.

How about this?

Not the tree... the lens. Notice the serial number of the lens!
The picture of the tree was made with that lens, attached to Canon 5D-II with aperture fully open, at 2.8.
Notice the bokeh!