Is ST Cube IDE a piece of buggy crap?

[peter-h]

That site identifies my CPU as 406C4 which is definitely is not...

[DavidAlfa]

  That site identifies NOTHING, it's a sample picture...

[paulca]

PHP is the VHS tape.  Revolutionary.  Flawed.  Popular.  Soon to be extinct.

Modern web, ala Web2.0 and beyond.

JavaScript primarily.  Most of what PHP et. al. did in the 2000s, now gets done in JS in the browser.  In many cases it also does the "Backend", even the REST APIs and database integration.

In a typical cookie cutter website these days:

Angular or React JavaScript "Single page application"
Containerised Microservice REST APIs providing data and processing requests from the front end. In order of popularity, sorta, Python, Javascript, Java, everything else including "serverless/codeless" cloud UI managed services.

Now sub-divide, layer, multiply out each and distribute globally.

[peter-h]

PHP is still the only server-side language which is maintainable into the future, without getting shafted with "modern" stuff like Ruby which is already dead and now costs 1.5k/day for someone to look at it. All these new paradigms will be dead in 10 years' time but PHP will carry on. I specify PHP for all server-side stuff.

JS is fine for client-side stuff but only to make things pretty; the functionality is wide open and anybody can go straight in, bypassing it.

What this has to do with Cube, I don't know That is written in Java and even STM don't want to touch it (e.g. to fix the famous random file opening issue). Java has turned out to be largely a waste of time, with as much runtime version dependency as a windows executable has on DLLs But an awful lot of people have put bread on the table at home with Java... not as much as with Ruby, which is still less good than COBOL-66

[zapta]

What is needed is an STM extension for Visual Source Code, that adds the STM specific functionality and leaves the rest to VSC.

Platformio is a good example.

[peter-h]

Do you mean using Microsoft Visual C++ development kit but for STM embedded instead of the 80x86 PC?

AIUI you can set that up if you know how. The VC++ kit can be used for any target. After all, these tools are mainly makefile generators, with debugging add-ons, and a file editor

[dietert1]

https://community.st.com/t5/stm32-mcus/how-to-use-vs-code-with-stm32-microcontrollers/ta-p/742589
Seems like VScode exists for linux as well.

Regards, Dieter

[zapta]

Do you mean using Microsoft Visual C++ development kit but for STM embedded instead of the 80x86 PC?

I mean using Visual Studio Code as the IDE and just providing the STM specific functionality. Compiler can stay the same as the IDE (gcc?).

[paulca]

PHP is still the only server-side language which is maintainable into the future, without getting shafted with "modern" stuff like Ruby which is already dead and now costs 1.5k/day for someone to look at it. All these new paradigms will be dead in 10 years' time but PHP will carry on. I specify PHP for all server-side stuff.

JS is fine for client-side stuff but only to make things pretty; the functionality is wide open and anybody can go straight in, bypassing it.

What this has to do with Cube, I don't know That is written in Java and even STM don't want to touch it (e.g. to fix the famous random file opening issue). Java has turned out to be largely a waste of time, with as much runtime version dependency as a windows executable has on DLLs But an awful lot of people have put bread on the table at home with Java... not as much as with Ruby, which is still less good than COBOL-66

If you wrote this in 2005 it might have been a little true.

It's 2025.  All of what you just wrote is seriously out of date.

* We don't have webservers anymore.
* There is no "pre-processor"
* Web sites run in JavaScript in the browser.
* The backend are JSON based REST API micro-services, which can be any language you want
* Javascript is not a "visual layer" technology today.  It's an application language.  It's natively compiled and executed directly on the CPU.  (Chrome and others wrap it in a sandbox and others go further and chroot it or vm it.  Browser Javascript execution was for example subject to the Spectre CPU cache exploit a few years back).
* Javascript is already very popular server side.  Search for NodeJS for example.
* I hate it with a passion.

This forum, in web terms if about 85 years old.  It's "old generation".  Web1.0. Server-side HTML generation, that whole pattern is dead, nobody above SOHO uses that model anymore.

Java's compatibility with itself is because the byte code version used by the compiler needs to match teh byte code version used by the runtime.  This is the same for nearly all "Virtual machine/managed" languages, like .Net, Java etc. etc.  The "cross platform: run anywhere" sell of Java says nothing about running a Java app compiled in a Java9 JDK running on a Java8 JRE.  It just says that a Java app compiled for Java 8 will run on any Java 8 JVM, regardless of it's platform, linux, unix, windows etc.

In a large part it is why Java struggled in the Desktop space.  That and the exceedingly tedious GUI and Widget toolkits and the horrible "Motif" theme it would default to.  Java Applets were an absolute nightmare.  Getting Java to work in a browser in 2002 was painful enough.  Then you learnt that there are 3 different incompatible JavaApplet API specs... which each need a different JVM?  F.. that.  They died.

Cube's bugs are cubes bugs.  Eclipse works just fine.  Java works just fine.  The CDT plugins and the STM plugins and scripts?  That a different thing.  Honestly when I was a C++ dev, I did have Eclipse CDT installed, but I would rarely use it.  It was great for discovery and better than grep, but the amount of actual "Eclipse" functionality, when compared to it doing Java, is night and day.  CDT has come a LONG way since then, but it's still clunky.

[NorthGuy]

This forum, in web terms if about 85 years old.  It's "old generation".  Web1.0. Server-side HTML generation, that whole pattern is dead, nobody above SOHO uses that model anymore.

It works. It works fast, in any browser, without bugs.

On most other sites I see happy faces, rolling circles, bugs, extensive spying, annoying pop-ups, hard-to-find content. Also most of them are very slow despite the Internet connection being over 1000 times faster than it was 25 years ago.

[peter-h]

This explains a lot of Cube IDE breakpoint issues - well done wek !
http://efton.sk/STM32/gotcha/g258.html

[rsjsouza]

This forum, in web terms if about 85 years old.  It's "old generation".  Web1.0. Server-side HTML generation, that whole pattern is dead, nobody above SOHO uses that model anymore.

It works. It works fast, in any browser, without bugs.

On most other sites I see happy faces, rolling circles, bugs, extensive spying, annoying pop-ups, hard-to-find content. Also most of them are very slow despite the Internet connection being over 1000 times faster than it was 25 years ago.

Precisely. For example, the AAC forum is much worse after the upgrade. I extensively use other fora for professional work and all them were much worsened after the "Web 2.0 upgrade" and their slow and wasteful interfaces (w.r.t. screen realstate).

This forum just works, and that is one of its strenghts.

[paulca]

This forum, in web terms if about 85 years old.  It's "old generation".  Web1.0. Server-side HTML generation, that whole pattern is dead, nobody above SOHO uses that model anymore.

It works. It works fast, in any browser, without bugs.

On most other sites I see happy faces, rolling circles, bugs, extensive spying, annoying pop-ups, hard-to-find content. Also most of them are very slow despite the Internet connection being over 1000 times faster than it was 25 years ago.

Precisely. For example, the AAC forum is much worse after the upgrade. I extensively use other fora for professional work and all them were much worsened after the "Web 2.0 upgrade" and their slow and wasteful interfaces (w.r.t. screen realstate).

This forum just works, and that is one of its strenghts.

Don't look at me.  I didn't do it.  I hate it.  I want my back button to work etc.

I'm just telling you how it is.  Leading edge, enterprise web is 90% Web2.0 model with Javascript client side applications hitting REST requests in the back end.

If you have no idea what I'm talking about, open "amazon.com" and go into the developer tab and check out first "Network" and see how many and how much is being communicated back and forth and also "Sources" to see the full set of application code.  It's about 4Mb.

[zapta]

It works. It works fast, in any browser, without bugs.

Not really. For example, I can't copy/paste images while writing a post.

Regardless of the technology stack, the user interface is tedious and dated.

[peter-h]

Images are always a problem. I run a tech forum (not electronics) and out of the 160GB server space, 112GB is used and of that about 60GB is user-contributed images. We actually spent some effort on keeping a lid on that, because most people are clueless and happily post a 20MB image, so we downsize stuff, etc. But still we have 60GB of it, after 13 years of running. Quality hosting (not some shit outfit like AWS) costs $100/month and you don't want to keep doubling and doubling that.

Also image upload is heavily browser dependent. You can throw in a library like dropzone but have to keep updating it. On another site I run we did it using basic HTML commands and it runs ok with Chrome, FF but not with some versions of Safari.

There are really no newer and better forum packages. The online community universe has moved on. If you want to create a forum into which people toss in banal one-liners (facebook, twatter, instagram sort of thing) then Discourse is OK but that was done in Ruby so again not a good way to go into the future. This forum package is good for tech subjects and detailed posts. And it probably doesn't get successfully attacked. There are many machines out there which run queries designed to exploit database structures and specifically generate slow queries, to eat CPU time. I am blocking one Hetzner /24 subnet after another because of these attacks. Extra indexes also help of course.

Advertising gets mostly adblocked so the revenue is not great.

[zapta]

Images are always a problem. I run a tech forum (not electronics) and out of the 160GB server space, 112GB is used and of that about 60GB is user-contributed images.

Do you mean that a tedious image upload user experience is good at reducing the disk space?  ;-)

Also image upload is heavily browser dependent.

Other modern web application solved this problem and you can drag or paste images in all of them with no problem. Most likely there are open source web components that already solved the problem and modern wed developers can use.

It's pretty obvious that this forum software is tedious and dated in today's standard.  I understand that there may be a risk and cost in upgrading but it's still tedious and dated.

[peter-h]

Sure they are "modern" ways but they also regularly break, and "somebody" has to go and fix it.

Regularly.

And has to be paid.

All IT systems regularly break and need fixing. It's just a way of life. And all online communities (except huge ones e.g. Mumsnet, or some huge US forums) struggle with the constant maintenance versus declining advertising income.

Finally, no forum package is backwards compatible. There is an army of "porting consultants" who charge a packet for converting a database from one forum to another. Usually they drop a load of stuff on the way across... so moving forward to not easy, even if you could.

You would not believe how much code there is behind something like EEVBLOG.

[zapta]

You would not believe how much code there is behind something like EEVBLOG.

I understand the cost and risk in upgrading, but we should also not deny that this is a tedious and dated system and that the gap is growing every year. 

"If You’re Not Moving Forward, You’re Moving Backward"

[peter-h]

The obvious answer is: wanna pay?

[NorthGuy]

I understand the cost and risk in upgrading, but we should also not deny that this is a tedious and dated system and that the gap is growing every year.

It's functional, well organized, easy to use, fast. I like it and it would be sad for me if it was replaced with modern crap. There are thousands of forums built as you like. There must be a few for me too

"If You’re Not Moving Forward, You’re Moving Backward"

If only we knew what is forward.

[paulca]

It works. It works fast, in any browser, without bugs.

Not really. For example, I can't copy/paste images while writing a post.

Regardless of the technology stack, the user interface is tedious and dated.

This.  It is weak in comparison to even a light weight blog platform these day.

However, look at the audience.

Hands up who is on Windows 7?

Hands up who is on an earlier version of Windows?

Someone mentioned speed.

Lets pick a use case example.   Searching.  When you search on this forum you complete the HTML form element up the top right.  When you hit Search, a CGI GET request is sent to the server.  The browser 'freezes'  The page reloads.  You need to download the WHOLE page again, even though most of it is the same templated repetitive HTML garbage.

In a Web2.0 version the page doesn't move.  The initial user-interaction is handled locally.  The search results view is rendered locally while in the background a REST (or REST like) request will be made to the search service.  it will responds with a few Kilobytes of JSON.  NOT massive verbose duplicative HTML.  The local JavaScript app will then use that JSON to generate the HTML locally.  Thats the "level 0".  Beyond that you can start to get clever.  Watch the user scroll the page, send your "future" for your next search result page and cache it with the expectation the user will request it soon.  Level 1.  What?  Why does the user need to select the next page at all?  If they have scrolled near the end of the results, lets just fetch the next page and render it without needing them to do anything - Level 2.  Amazon, facebook and YouTube are at Level 104.  That opens the privacy, monitoring, data-collection, user profiling thorny issues, but as you guys know, the tech and how its used are not the same kettle of fish.

For those of us who have modern hardware made in the last 5 years, this happens next to instantaneously.  Certainly the local side of it.  The REST request should take a few dozen milliseconds, but depends on load and your internet connection.  However it will be faster than a PHP script, by orders of magnitude in many cases.  It will also use a magnitude less resources and cost a magnitude less to host. 

You have a 8/16 core CPU and its ONLY YOU using it to generate HTML.  On this forum, it's you and ALL currently active users, bots, search engines etc. competing for the resources, while Dave pays for it.

Both models have pros and cons and there are situations where one or the other would be inappropriate.  Most situations I think benefit from the more modern approach.  Not just from the "User's POV", but from the business and development point of view.

Computers get faster.  Humans don't.
Computers get cheaper.  Humans don't.

This is why enterprise software is not like MCU development.  Our landscape evolves orders of magnitude faster as it has been challenged with complexities, scales and scopes orders of magnitude greater.  When you divide and conqueror, you will always create waste and slack.  If those are all you can see though, you are blind to the bigger picture and to the scale involved.

[paulca]

Images are always a problem. I run a tech forum (not electronics) and out of the 160GB server space, 112GB is used and of that about 60GB is user-contributed images. We actually spent some effort on keeping a lid on that, because most people are clueless and happily post a 20MB image, so we downsize stuff, etc. But still we have 60GB of it, after 13 years of running. Quality hosting (not some shit outfit like AWS) costs $100/month and you don't want to keep doubling and doubling that.

Different world, different scale, different era.

Have you costed an S3 bucket object store for example?  For 60Gb it shouldn't cost you more than about $2.50 a month rental.  Traffic to different regions can be pre-negotiated and it's not that bad.  Not compared to buying and installing a larger disk is.

At the level I work at, 160Gb is perfectly reasonable size for an in-memory cache before the request even gets to the server.

The storage would NEVER be physical disk.  It would be an elastic allocation on a SAN.

My home dev server has 128Gb of RAM and 18Tb of storage. 

The last PSI (Physical server instance) I logged into, in a bank, was a basic 'dev' server for trialing software on.  It was an HA hot-hot pair.  Behind them sat a pair of paired NAS drives.  So dev1 and dev2 where mirrors, each had /data and /data1 which where individual stores but also mirrored across 2 hosts in the SAN logical arch.  Each instance "sees" 512Gb of RAM, 196cores and /data and /data1 are 2Tb. 

We used to get nasty emails from devops regarding leaving 100Gb files in /tmp as /tmp was a RAM disk.

[zapta]

However, look at the audience.

Hands up who is on Windows 7?

My windows 10 computer tells me that it will stop getting security updates in October.

[NorthGuy]

it will responds with a few Kilobytes of JSON.  NOT massive verbose duplicative HTML.

Give me a break. A very big forum HTML page is about 100k which compresses to about 20k. At 200 Mbit/s connection it's less than 1 millisecond. How's that a problem?

A java script to parse JSON wouldn't be too bad neither, but, for some reason, you end up downloading megabytes of different scripts from various Web sites, possibly wrecking the whole thing if one of these source sites is down. Is that all to minimize the size of the downloads? Looks like the cure is worse than the disease.

Why does the user need to select the next page at all?  If they have scrolled near the end of the results, lets just fetch the next page and render it without needing them to do anything

And then everything unexpectedly jumps and the user no longer knows where in the list he is.

[elektryk]

And then everything unexpectedly jumps and the user no longer knows where in the list he is.

Imho those endless scroll approaches are often crap, forget about ctrl+f, forget about URLs to exact place...