Is ST Cube IDE a piece of buggy crap?

[eutectique]

Yesterday I've stumbled across the discussion on reddit named Interested in improving CubeMX, or in venting about it? We're searching users to talk about UX and started recently by a Lead UX designer for the CubeMX team in STMicroelectronics.

I don't have an account, otherwise I would steer the guy to the bottomless well of inspiration here

Anyone?

[peter-h]

That is Cube MX



MX - the "click click click code generator" is a whole other debate.

My project started with MX (before I was involved with it and while I still believed the guy who started it was actually going to ever finish it) and used it to generate lots of code fragments to get quickly started. I've written all this before but yeah it gets you quickly started but then you get into lots of dead ends, end up googling like crazy for fixes, and in the end, months later, you have something working. Overall, yeah, it saved you time.

I took over that project c. 2021 because it became obvious it would never be finished. It is now finished but jesus what a lot of work. MX generates code for stuff like ETH low level (the famous "low level input" etc code) and LWIP integration but there are so many gotchas there...

If you are doing something simple like ADC -> some code  -> DAC and loop back to start, then MX is a good timesaver. But most modern products contain a ton of connectivity and that stuff is packed with gotchas.

I used to have a reddit account but they deleted it for some reason. Also it is structurally a really crappy forum, for intelligent discussion. It is for "one-line tosser" type of posting. Like Twatter (sorry, X).

[NorthGuy]

It ended up in the mess of the 1980s which nobody wants to go back and fix, get some of it still exists.

Give me a break. It was the same in 80-ies. It was the same before then (read the description of Algol-68 for example). Programmers had, have, and will have OCD. That's in their very nature. That's why they wanted to be programmers in the first place. So, they try to make everything universal, structural, systemic, reusable. Then they use all these things which are swarming in their brains and try to shape reality to match their picture of the world, while reality resists. When they're not satisfied with the results they try to add more systemic stuff on top. Just like drug addicts, gamblers, or people in deep debt, believe that the solution to their problem is to use more drugs, gamble more, or borrow more money. And convincing them otherwise is utterly hopeless.

Instead, they need to strive for simplicity. Don't add complexity where it is not needed. This works just as Occam's razor in science. If you strive for simplicity, other things will fall in place automatically. One programmer said: "Novices ignore complexity. Professionals deal with complexity. Geniuses remove complexity." That's the best programming advice I've ever heard.

[peter-h]

Great last sentence

they need to strive for simplicity

But, that directly contradicts putting bread on the table at home.

It's obvious that the vast majority of people who post meaningful stuff here are working for themselves. They are not hourly paid employees, who are only too happy with a complicated solution. If you work for yourself, the only way you will ever make money is by recycling experience. I used to do consultancy but now just run my own design+mfg business.

[westfw]

Code: [Select]

  DigitalWrite( D1, UP );
  DigitalWrite( D1, DOWN );The frustrating part is when "the community" implements digitalWriteFast(), and the vendor ... ignores it.

Although there was an interesting discussion recently along the lines of:

Why is

Code: [Select]

   digitalWriteFast(2,HIGH);
   digitalWriteFast(3,HIGH);
   digitalWriteFast(4,HIGH);
   digitalWriteFast(5,HIGH);
   digitalWriteFast(6,HIGH);
   digitalWriteFast(7,HIGH);
   digitalWriteFast(8,HIGH);
   digitalWriteFast(9,HIGH);

So much faster than:

Code: [Select]

   for (i=2; i<=9; i++) {
     digitalWriteFast(i, HIGH);
   }


[NorthGuy]

I used to do consultancy but now just run my own design+mfg business.

How does this force you to choose complexity over simplicity?

[DavidAlfa]

Why does every programming/MCU thread eventually get taken by OCD people?
This thread is not about programming practices, not about 80s was better, not for "I use emacs and makefiles", "Linux better", "Ahh this fkcn crap", "there's only one way - my way, you're all wrong".

This is for CubeIDE!
- How to do xyz in CubeIDE.
- How to fix xyz in CubeIDE.
- Bugs in CubeIDE.
- News in CubeIDE.
- Whatever - in CubeIDE.

If you don't like it, don't use it, don't post here how much you hate it every 5 seconds.

[eutectique]

Is it stackexchange?

[SiliconWizard]

If you don't like it, don't use it, don't post here how much you hate it every 5 seconds.

That was the very topic of this thread.

[peter-h]

How does this force you to choose complexity over simplicity?

It doesn't - I still go for a simple way every time.

Why does every programming/MCU thread eventually get taken by OCD people?

Because a lot of people on forums have a lot of time

I see Cube IDE 1.16.0 is out. So, give it a month or two for 1.16.1 but that will still very likely break a load of stuff in your project. Why doesn't ST fix the most obvious annoying bug: the "pseudo-random" new file opening? I can't see a serious dev installing every new version when it comes out. I am stopping at 1.14.1...

[wek]

Yesterday I've stumbled across the discussion on reddit named Interested in improving CubeMX, or in venting about it? We're searching users to talk about UX and started recently by a Lead UX designer for the CubeMX team in STMicroelectronics.

It's telling, that they don't start such discussion on the CubeMX-specialized part of their own forum.

Oh, and it's anonymous, too.

JW

[peter-h]

It is a risky thing for that guy to be doing. I mean... there is a thing called google This suggests multiple things

- stupidity
- knowledge the product has problems, and a desire to not publicise them on ST's own site
- avoidance of ST's employee social medial policy
- stupidity

[dietert1]

Whenever brands are attributed positive or negative in the public (words like "crap") we may assume someone paid for that. Of course we also have some notorious "worse than expected" contributors, just bad mood.
Although general wisdom tells the opposite: i prefer learning from success stories. I mean contributions explaining how something worked better than expected. Why lament the same issues over and over again?

Regards, Dieter

[peter-h]

Indeed, and I have posted the good things about Cube IDE, when others suggested commercial products which are sure to bite you in the bum one day with the licensing enforcement. Cube does work overall and probably as well as commercial products whose issues do not get posted openly.

[wek]

... commercial products which are sure to bite you in the bum one day with the licensing enforcement.

Not every producer plays the games of continued money sucking. There may be such who still perform the good old plain straightforward single-shot direct exchange of product and money.

How the product performs, what are the guarantees, and whether repairs are available - for free or for money - is a whole bunch of other questions.

Cube does work overall and probably as well as commercial products whose issues do not get posted openly.

Okay so can rephrase the opening question to be "is ST Cube IDE the same piece of buggy crap as any other IDE out there"? ;-)

JW

[peter-h]

There may be such who still perform the good old plain straightforward single-shot direct exchange of product and money.

Sure, although in "CAD/EDA" this is nowadays unusual. I know someone who uses a "modern" PCB design program which he rents for about 500/year. And I would put money on it not opening a design done 10+ years ago Well, not without some, ahem, "editing".

For the product I did, there will be an option for customers to write own code. That I accept is an unusual dimension. I have been selling such products for ~30 years and it's become obvious that there is almost total resistance to paying anything for the toolkit. I used to sell a C compiler for £450... the other day some guy in Germany phoned and refused to pay for that, saying he can get Arduino tools (or maybe some other kit; I have not followed the kit business closely) for nothing. Whether that option was any good for his application was irrelevant.

But the above creates another challenge: I have to freeze the product at some version of Cube IDE. It is obviously impossible to chase after ST's endless new release stream. They break stuff all the time, change stuff in a way which creates compiler warnings/errors, and I will make money only once when I sell this thing. So I will have to keep 1.14.1 on a website It does have the signature on the .exe so customers should trust it OK.

Okay so can rephrase the opening question to be "is ST Cube IDE the same piece of buggy crap as any other IDE out there"? ;-)

I don't think one can edit the thread subject

[paulca]

It ended up in the mess of the 1980s which nobody wants to go back and fix, get some of it still exists.

Give me a break. It was the same in 80-ies. It was the same before then (read the description of Algol-68 for example). Programmers had, have, and will have OCD. That's in their very nature. That's why they wanted to be programmers in the first place. So, they try to make everything universal, structural, systemic, reusable. Then they use all these things which are swarming in their brains and try to shape reality to match their picture of the world, while reality resists. When they're not satisfied with the results they try to add more systemic stuff on top. Just like drug addicts, gamblers, or people in deep debt, believe that the solution to their problem is to use more drugs, gamble more, or borrow more money. And convincing them otherwise is utterly hopeless.

Instead, they need to strive for simplicity. Don't add complexity where it is not needed. This works just as Occam's razor in science. If you strive for simplicity, other things will fall in place automatically. One programmer said: "Novices ignore complexity. Professionals deal with complexity. Geniuses remove complexity." That's the best programming advice I've ever heard.

Absolutely.  What do you do when you have more than one programmer or more than one team of programmers?

Programming != Software Engineering.

Take an example we are all familiar with.  This forum.  While dated it's still a substancial bit of software.  If you were to go and investigate what exactly is involved from a "near baremetal" point of view you would probably require the rest of your life to do so and you still won't be an expert in much of it.

Put another way.  If someone in their 20s were to start out with a bare-metal computer and write every line of code required to recreate a forum like this... they would not be able to complete it in their life time.  Not to the same standards as it is here today.

Even if we take the largest component out, the RDBMS, and the second, the webserver, we are still left with something too large for one person to implement inside of a decade.

So what do we do?  Do we hire a bunch of "coders" "programmers" and just let them loose?  NO!  There lies dragons.

What frustrates me with MCU dev is that basically it's solving the same god damn problems over and over and over again.  MCU's are ulimtately simple.  Yes they are full of detailed fiddly bits, but it just annoys me that these cannot be encapsulated into a reusable form.  Arduino is about the closest.

It's not like every single project is 100% despoke and unique.  Seriously look at the total code you have written for MCU and honestly say that at least 60% of it is duplicated from previous projects.

The challenge in writing such components and having them fit together in thousands of combinations is not easy.  Maintaining the small foot print and performance even harder, but it should not be impossible.  Sacrifices may need to be made.  It will most likely not be a "one size fits all".  Some frameworks will be better at somethings than others.

I think that cost is the biggest hurdle.  With MCUs normally requiring software which is sold with the product and never independant of it the margins are very tight.  If one or two MCU devs can write almost bug free and almost useful firmware for $40k, why would you invest in a senior software engineer to facilitate faster, more efficient development if it costs another $40k?

EDIT:
Probably the largest amount of work in terms of enterprise software is not in writing it, the first time.  It's in dealing with requirements changing constantly.  Managing these.  Modifying existing software to deliver to these changing requirements.

With MCU code being so small the temptation is always to start from scratch and "do it better this time".  However, if you are releasing new updates to your firmware on a weekly basis and your business requirements are changing weekly.  Good luck.

So in Big Iron we don't just encapsulation and engineer out repeatitive code we also factor in flexibility, extensibility and change, state, transition isolation, isolation of concern/responsibility, such that when one component changes the others will continue to work or error out sensibly.

[peter-h]

Well, Cube IDE is a piece of junk







This is not a new file, either. It's been there for years. Something has happened to the file system. But I can open it in Cube, and of course in Windows. The other .ini files work OK.

There is a fair amount on google about these cases but only talking about paths for different filetypes (e.g. .cpp) being set up wrong. I never set up any paths for .ini and the other files have always worked.

The issue seems to have happened when I created another .ini file.

It is known that Cube IDE likes to have a Clean Project and Index Rebuild after any addition of files.

[dietert1]

Reducing the search path for each file type may be a valid attempt to increase performance. Except the feature needs configuration. Your problem with a new file may be caused by some cache memorizing the location of known project files - another attempt to increase performance, hiding the bad configuration until now.

Regards, Dieter

[Siwastaja]

Totally unrelated, but .ini name extension for something which clearly is in C syntax (probably declaring/defining variables, maybe preprocessor stuff) is a pretty weird choice, and would confuse others, because .ini is well known extension for Windows plain text configuration files since early 1990's if not earlier. I wouldn't be too surprised for this to confuse some IDEs / code editors, too.

Just use default_config.h like everybody else does.

[paulca]

Just use default_config.h like everybody else does.

"Convention wins over process"

The next step towards happiness is to remove boiler plate by setting all config and init state to the most common required configuration by default.  Then you only need config for your bespoke differences.  We call this "Convention over configuration".  It can produce a bit of a feedback loop.  Framework developers steer you towards the most "normal and useful" config because it's 100% defaulted.   So those who didn't know what the conventions and best practices were, will get steered that way now.  So it results in more people doing things in the conventional way, which leads to more code reuse and a less fragmented engineering space.  It's only if you need bespoke config and know what you are doing that you need to set any.

[Nominal Animal]

Okay so can rephrase the opening question to be "is ST Cube IDE the same piece of buggy crap as any other IDE out there"? ;-)

I don't think one can edit the thread subject

You can.  Just edit the initial post in the thread, specifically the title in that initial post, and it will change the thread subject (or topic) seen in the topic lists.

I did so in this thread.  The original title was "Why not use digital isolators?" as you can see in the first few responses (they retain the original title used), and I changed it to the more specific "Why not use digital isolators in hobby microcontroller and USB designs?".

[DavidAlfa]

You still want a header to easily config everything, instead modifying the default values in 50 files.
C'mon this is basic stuff. Using a non-standard file extension doesn't make any sense, ini doesn't have any place in C.
Another rant against CubeIDE where the fault is in the programmer!

[peter-h]

That "no such file" error has magically gone away. Yes, a probable caching issue.

BTW I have been using .ini file(s) in this project for about 5 years. It would be weird if this was silently failing.

[SiliconWizard]

Okay so can rephrase the opening question to be "is ST Cube IDE the same piece of buggy crap as any other IDE out there"? ;-)

I don't think one can edit the thread subject

You can.  Just edit the initial post in the thread, specifically the title in that initial post, and it will change the thread subject (or topic) seen in the topic lists.

I did so in this thread.  The original title was "Why not use digital isolators?" as you can see in the first few responses (they retain the original title used), and I changed it to the more specific "Why not use digital isolators in hobby microcontroller and USB designs?".

But all the "Re:" replies will still show the old subject, won't they? So...