Posted on October 11, 2016 at 12:24> When speaking about ''Cube mess'' ...
I spoke about my personal impressions, both with the package, and with other people's problems manifesting here.
I have the following points:
1. The ''Wizard'' approach. Letting inexperienced (ignorant ?) people create an application with a few mouse click seems o.k. at first glance.
But not only fails this approach (often miserably) when the desired goal deviates from the use cases imagined by the designer. (Do you remember the MS-Word/Excel 'Clip Wizards' ? Where did they go ?)
It also fosters an ignorance of the underlying subject. Patching an app together via a few click does not make one understand anything. It might even keep one from understanding it.
As another famous poster expressed it ''Monkey press button software''. I guess it not only incidentally correlates with the flood of semi-literate engineers on MCU fora.
2. The ''All embracing'' approach. Having used the SPL for many projects, I for sure tried Cube.
But where the SPL consists of quite autonomous modules, Cube is helplessly intermingled, and enforces all it's concepts on applications using it. Trying to remove Cube's use of timer, callbacks or just it's include file dependencies is a PITA. I resort to direct-register-access, or to MCUs for which a SPL is available.
3. Seems you cannot keep up. The cube developers are obviously overwhelmed by the amount of work to do for new MCUs.
And with more complex variants (like the M7 and high level M4), the possible inter-peripherals dependencies and use-cases grow exponentially.
Where are the LL (low-level) cube headers for most MCUs ?
I currently use STM32 products only for my private projects, so you can safely ignore my (very personal) opinions.
But I'm not sure if the Cube approach is heading toward the right direction.
Apart from that, I think Liviu did a great job, and I'm glad you finally managed to add it to the list ...
