Posted on August 17, 2017 at 14:38Since there's no real documentation of their implementation, and importing their code into an existing project is going to vary widely between projects, compilers, and even development environments (I can't begin to imagine how eclipse would organize this, or anything), I'm not sure if certifying the source code would buy much; too much is left to implementation.
To obtain UL certification of the end product, it's nice to be able to say 'we used this code library which has its own certification' but from my perspective, the STL code (and the Cube-generated code) are black boxes of unknown reliability and need to be treated as such. It's nice that some of the lower-level UL safety concerns like memory corruption are handled by code I don't have to write, but from our perspective the software I do write (and the hardware it runs on) needs to handle as many types of faults as possible, because there are >100 source files that we didn't write running under our code. I view this as acceptable in this specific application, but I can easily imagine other applications where relying on auto-generated code would make me very nervous about human safety.
