ARM coders are nuts

Just spent an hour ploughing through the PRM’s regarding events, vectors, FIRQ’s and the like.
It’s really interesting reading, but I have to come to the conclusion anyone who likes this kind of thing is clinically insane.
A state I already know I’ve been in for many many years – perhaps I was born like that?
Time to watch some mind numbing TV to recover some sanity.

some sanity

I’m told it’s over rated. Spending my day doing network support and my spare time¹ frequently doing the same² I think I will just take other peoples word on it.

¹ Spare, spare time may sometimes involve being in a pub³ that has a selection of good ales with like minded people

² There’s an iPhone close by that hasn’t gone “bong” for a few hours, which is a bonus

³ Sadly almost invariably with the above mentioned iPhone

but I have to come to the conclusion anyone who likes this kind of thing is clinically insane.

:-(

On the flip side, how can anybody stand to be an “app” writer, coding for a virtual machine and a whimsical API with no real understanding of anything that is going on inside? If “year of code” is showing us the future, then god help us all.
There will always be a need for assembler level programmers because once you strip away all those JVMs and OOP rubbish….that’s what is left. The bare metal. The thought process that separates script kiddies from real (wo)men.

So nerr, etc.

A state I already know I’ve been in for many many years – perhaps I was born like that?

I think I was born that way. When I was young, my favourite word was “why?”. Not to annoy grown ups (though it frequently did) but because I actually wanted to know. Everything. All at once. I defeated my childproof car seat before my parents figured it out. I could barely walk before I took the record player and eight track apart (getting it back together was more difficult, but hey, quadrophonic sound was overrated anyway!). Fixing washing machines and VCRs was an after school job. Then I went to boarding school and met the Beeb and had entire evenings (and sometimes nights) of free time.
The rest is a slightly less interesting history, but I’ll be happy to embellish the good bits…invent a love interest or two (coders are people too!), add zombies, or ninjas, or hey – how about ninja zombies? The epic battle of one lone programmer saving the world from ninja zombies one broken keyboard at a time! ;-)

aaaaa

Well some people used to use machine code rather than assembly back in the day!

aaaa

there is not that much combination of opcode in the bottom, on the arm it’s even simpler.

Anatomically uncomfortable?

aaaaa

Oject oriented programming is very powerful, with assembly there is just a point where the time of development and debugging, fixing memory leaks and all that become too much time consuming, compared to object oriented programming, where lot of things can automated, and make debugging much easier.

Don’t get me wrong – I write a lot of code in C and BASIC because – as you say – writing and debugging pure assembler is a challenging process for which easier methods exist, provided you are willing to put trust in a compiler and somebody else’s library. We have reached the level where reinventing the wheel over and over and over is… well… a bit dumb.

However, I wonder how many modern programmers even understand the machine that they are using. Do they really understand how a computer works, or is their programming the equivalent of using VisualBasic? I have written numerous applications for Windows and I have no idea about how the Windows multitasker works. There was no need, I would just hook code to events (icon click, window close…) and VB would do the rest. The payback for this is all the baggage that is the VB runtime, plus the speed hit that VB applications incur – especially those compiled as P-code.

This is why I am glad to have grown up in the days of the BBC micro. Each trace on the circuit board, each logic gate… they all had a specific and clear purpose. You could understand the entire machine in a way that isn’t possible these days. You could improve your ability to program by being able to see the world from the processor’s point of view.

Instead, the common approach now is instead of improving the code, improve the hardware. Am I the only person who thinks it is a bit ridiculous to be running RISC OS on a gigahertz multicore processor with a gig of RAM? That’s… mindblowing. The thing is, we can eke out perfectly acceptable performance (say, with OvationPro) by running a single core at its slowest speed. Other operating systems need all that oomph just to get an acceptably responsive UI…

aaaaa

I had an interview a few years ago and the guy was amazed he met someone else who knew what context switching meant.

Sometime the way the computer works doesn’t matter that much, it’s just moving bit from memory

Yes, it still matters because of what is the most efficient on the platform in question. You could construct a word using four byte reads, but a word read is more effective. For loading multiple words, you could chain together twenty eight byte reads and some bit shifting, or you could use a single LDMxx. The problem is, if you are using an arbitrary virtual machine, you don’t necessarily know what it is doing under the hood.
I can give as a great example the VB5 runtime (and perhaps VB6) which had no bit shift operations. This meant that bad programmers would shift using meh = meh / 2 which would do a >> operation. Only with the side effects of prompting “meh” to a float, dividing, then casting “meh” back to an integer (it was better to use the ‘\’ division operator as that would int divide and just discard the remainder).
I bet you thought I was joking about repeated byte reads to load a set of words. Read the above again to try to understand what must have been going on to support something even the x86 could manage in a single operation.
By the same token, it is possible to write programs using nothing but “Variant” data type (or whatever the language equivalent is) but it comes at an expense of taking a lot more resources to get a simple job done. Okay, it might be useful to be able to assign a string value like “123” and have it be 123, but isn’t this what casting is for? Or atoi() etc? Is the ability to be lazy really worth the cost it will incur?

‘download this file slowly at medium preioriy’

I think that is my PC’s default setting. It’s something of an indictment when it is quicker to do stuff on a mobile phone… That said, I think as each month passes, the distinction between traditional “PC” and “tablet”-like is narrowing. You can get PCs (Lenovo?) that look and feel like tablets. Windows 8 seemed aimed more at them than real PCs.

often it’s used to hide some monstruosity that no sane programmer would touch with a ten foot pole,

Been there, seen it, ran away screaming.

and it provide some kind of syntax to manipulate the shmiilblik.

I have no idea what “shmiilblik” means (sounds like a Yiddish word), but I like it. Spaghetti code of the sort that poisons the mind of anybody that tries to understand it? Like the Sadako of programming? (click link if you didn’t get the reference, it’s a picture)

at first it can make you wonder if you’re too dumb to understand it or something,

Oh hell yeah. I have a copy of the XP API as a big searchable “book” file on the computer I was using at the time and sometimes when looking up stuff, or just reading random bits out of idle curiosity, I’d back up, reread it, and say a rude word in a slow emphasised Keanu Reeves kind of way.
You know the thing about monkeys and typewriters? I think the Win API was monkeys on acid.

the guy was amazed he met someone else who knew what context switching meant

Well, then. That pretty much makes my point.

However, I wonder how many modern programmers even understand the machine that they are using.

Sadly I do not…or at least only the very basics.

I see this as a weakness and something I plan to correct (which is why I asked for the Pi and some books on low level stuff for Christmas).

I have plied my trade for some years now producing GUI programs that are generally several libraries, frameworks and an operating system (or two) away from the hardware yet I feel that all I have managed to do is put on a pretty light show in the corner of a large and foggy room. Its impressive, sure, and serves a useful purpose (mostly) but I can’t help feeling that I am missing something fundamental – something that will make me a better programmer and clear away the fog.

Hopefully over the next few years I’ll get the time to do more study and increase my understanding but sadly I know far too many programmers who simply aren’t interested…

I had an interview a few years ago and the guy was amazed he met someone else who knew what context switching meant.

That’s surprising. My wife…

Boom, boom!

The thing is, we can eke out perfectly acceptable performance (say, with OvationPro) by running a single core at its slowest speed. Other operating systems need all that oomph just to get an acceptably responsive UI…

yeah, but just image RO using all the power of a multicore device

Hopefully over the next few years I’ll get the time to do more study and increase my understanding but sadly I know far too many programmers who simply aren’t interested…

While you learn you live, stop bothering to learn and you might as well stop breathing.

ARM coders are nuts?

Well, ARM was invented by a company called Acorn! ;)

aaaaa

aaaa

aaaaa

basically it’s something unknown that you have to figure out, and in the sketch colluche can never even pronounche the name correctly lol

Ah, you mean “the known unknowns”, right? ;-)
I only know Colluche as the face on the Resto de Coeur publicity.

it can be useful that some part of the program can be able to manipulate some data without having to know it’s type, and then passing it to other routine who can then figure the type of the data from the variant type.

I think this surely depends upon how you have specified the program? In C it is possible to use “void *” as a placeholder for “could be anything here”, though I have tried to avoid such things myself.

When you compare the netbios system design with dns system, make really wonder why microsoft did their netbios thing,

That was back in the days when Microsoft (infamously) figured this whole “internet” thing would never really take off, and it was surely much better for Microsoft to create their own proprietary protocol that they could control…

Back in those days, Microsoft’s business plan was to either buy small companies for their product, or to make their own to crush competitors. Look at the history of FoxPro. Look at WordPerfect/AmiPro vs Word. I would even say look at Netscape vs MSIE – Netscape was utterly crushed by IE but sensing the changing of the wind, they ultimately had the last laugh giving Netscape to the world from which Firefox was born, the browser of preference on Windows systems. IE, which was going to have its final definitive release with IE6 (which is why so much ActiveX stuff is tied to IE6, stupid devs actually believed that) has been playing catch-up since and, well, let’s just say that Microsoft isn’t so great with compatibility with niche things (which is why so much ActiveX stuff is tied to IE6, later versions won’t run it, or won’t run it correctly).
Basically, the early history of Internet on Windows is fifty percent Microsoft utterly misjudging things, and fifty percent Microsoft thinking it could throw its weight around and “own” whatever protocol people used. Sadly for Microsoft, the background of early internet adoption came from a Unix background and they had no truck whatsoever with that sort of thing.

The thing i have more and more the impression of also is that lot of the choice of design are also more impacted by a myriad of legal/commercial reason more than because they make sens from a technical point of view.

You can blame the messed up US patent system for that. It seems you can patent an existing idea by suffixing “on a computer” or “on a mobile phone” to the description. And, it also seems that patents are granted with.. not a lot of attention paid to “prior art”, seemingly with the idea of “let an expensive court battle sort it out”. Unfortunately this means that (within the US at least) programmers have to be very careful of what and how they implement as the large rich companies trade and collect patents to preserve their innovation use as a weapon. Look at Microsoft’s claims of patent infringements within Android, their collecting some money for each device sold by some manufacturers with very little disclosure of what exactly is concerned here. It may be valid, or it may be blackmail. The smaller manufacturers tend to prefer to pay up rather than get into a costly legal battle with a legal process that has already shown considerable bias for the home team (re. Apple vs Samsung).

For my own code, I don’t tend to pay much attention to this sort of thing. I figure that if I can think of something then it was never worthy of a patent in the first place, and anyway I’m a European so sod off… I’m not a company, so I have no assets to go after and I’m not a company so I have the flexibility to rewrite some of my website rules to block entire continents if it came to that.
However things like RISC OS… must be more careful.
For instance, there is a reason why the Long FileNames handling of FAT32FS is botched (LFN vs 8.3). This is by design and this is exactly for this reason.

how can they get to that extrem that there is a built in function to just burn the phone out to protect the firmware from being modified or read,

? Why bother burning out anything.

Some PICs and 8051 family with onboard flash have a special option that can be set that once a firmware has been programmed to the device, the option can be set (and never unset) which means that the firmware is then only available to the CPU and any attempts to read it will return gibberish.

I’m not sure it’s that much about being lazy, but it’s all about error resillance etc,

I’m afraid I would consider code that could accept 123 (as a value) and “123” (as a string) and consider them equal would reduce resilience and increase problems. PHP seems to be built upon this idea and I have hit errors in my own code where something have been interpreted in a way different to that which I expected.
To clarify, the “lazy” comment was referring specifically to the use of a variant data type instead of using proper data types and casting/converting (if necessary) between them.

i know by just checking some input for null pointers and doing some sanity check on some pointer can just keep the thing going and will have no impact on the execution,

That is an entirely different thing. I think a lot of the zero days and privilege escalations are due to a basic failure to sanitise input. If you write a module (generic module, not a RISC OS module!) to do “something”, then you should never assume that those passing data to your code even bothered to read the API, that they didn’t litter parts of the data with their own magic values, and that they themselves haven’t failed and will be providing bad data (like trying to read past the end of the file but not trapping it because “oh, that will never happen in production code” – famous last words…).

You shouldn’t go over the top, there comes a point when the input should just be faulted, but then a module shouldn’t bring down the system because of common case screwups like null pointers.

It’s going this route that they implement layers and layers of things to check any input and have status, error code, and everything to avoid crashes

…would be nice, then, if it worked.

instead of fixing all bugs to begin with, they prefer to do simple error checking

I’m afraid the mentality is much worse than that.
In the past, a lot of people have criticised RISC OS’s rather stark lack of catching errors. It is possible (perhaps even easy) for a badly written program to bring the system down. Okay, granted, our process of running “module” (RISC OS module now!) code in what is effectively kernel space is… maybe not the best idea post Y2K, but we’re stuck with it.

HOWEVER, if you flip the argument around, you can ask a very valid question: So as a programmer, one is absolutely depending upon the OS to catch mistakes? What kind of programmer is that?

In linux they love to cast ‘opaque structures’ to void pointers,

I noticed. Ugh.

aaaa

many people would easily say that the flash plugin running in firefox is an activeX whereas it’s built on an entierely different technology

Code to support a feature currently in wide use. Has known vulnerabilities and is now being replaced by an alternative – various people dropping flash in favour of HTML5 use.
In a similar vein Java is a pain in the nether¹ and has so many security holes even the USA government is kicking Oracle. Why people haven’t already ditched it and coded with something without the faults is the real question. It’s not like the vulnerabilities first appeared recently.

I would rather use a static ip than a netbios name personally

Manually entered or DHCP assigned? :)

¹ Ever wondered why organisations like the NHS still hang on to XP and old versions of IE? Some plonkers coded in a way that relied on the security holes in order to work. Anything with the holes fixed stops the business critical applications working. You check for alternate applications and they usually have even older dependencies with bigger holes.

aaaa

aaaa