The F-35's avionics are mostly in C++. The F-22 was the last fighter program to use Ada heavily AFAIK.
You are correct. But the Ariane crash was the result of improper testing when reusing (valid and correct) code in a new situation that resulted in the crash.
Better testing and development procedures would have, potentially, prevented this. And it could have occurred regardless of the languages involved.
What Ada does bring to the table, however, over C, C++ and Fortran (the other 3 primary languages used in the embedded avionics world) is a much better type system and concurrency system that gives greater confidence when developing the system. Much as the ML worlds type system reduces or eliminates certain categories of errors, or moves them to compile time rather than runtime. Where they are still detected in runtime, Ada also offers, particularly during test and development, much greater ability to determine the earliest location of the error if the type system is being exercised properly.
No.
Here's a timeline of the plane so far [0] and a page that includes some details from 2014 including a master schedule and a production status [1]. The 2050 comment was presumably an exaggeration referencing the frequent problems and delays with the development.
[0] http://aviationweek.com/F-35Milestones [1] http://intercepts.defensenews.com/2014/09/the-current-status...
https://stackoverflow.com/questions/9827176/what-is-the-pred...
I'm not sure there is anything I'd build with such insecure languages in 2016, let alone an aircraft or flight system.
I think it's a gross oversimplification to state that the avionics software is developed in C/C++: the avionics software is using a subset of C and/or C++, along with support tools, system engineering models, code reviews, V&V, etc. As I see it, there isn't that much difference on this sort of project if one were to use Ada or Java instead of C or C++: the support tools may be somewhat simpler, but the overwhelming majority of the SDLC activity remains unchanged.
I'm not clear what causes this radar system defect, I didn't see it described in the source article, but a defect in design (for example) would seem likely to occur regardless of the programming language selected.
Also, I'm curious to hear what your proposed alternative would be and why?
There isn't just one CPU in this aircraft. There are probably dozens or hundreds. Not all of them can be powerful without undesirable tradeoffs.
That being the case, I wish I wasn't too late to edit my comment to be a tad less snarky.
That said, I understand a bit of the trade-offs involved (though far from an expert). I think there's room for discussion somewhere in the middle. The parent took one extreme ("you had it coming using C/C++"), and at first reading your comment stated the opposite ("C/C++ is just what you use on embedded").
Knowing very little on the subject at hand (embedded avionics SW), I'd still argue that maybe there at least be a discussion around using something like Ada for critical systems? (Yeah, I know what a big hit Ada is with embedded folk, at least the ones I hang around.)
Question: Languages used to develop software in previous project (multiple answer)
Ada = 28.9%, C = 48.9%
Question: Languages used to develop software in current project
Ada = 12.2%, C = 53.3%
Since then, in other jobs in the same general field (avionics or related), the answer has been the same at most offices.
Go ahead and actually support why C/C++ produces more verifiable and secure code than Ada?
The US military doesn't develop software. They also have very little choice over who does. The entire system is a convoluted mess of lowest bidders and inept contractors.
So it is ultimately up to the lowest bidder what they want to develop in. There just isn't enough incentive to develop high quality, low in bugs, software for them.
It's a painful and expensive process, but it can work, as demonstrated by the handful of teams that have reached the ~1 bug per 1e6 opportunities defect rate _all_ using variations of the above (tho some of them use Ada).
Let's check ourselves a bit and realize developing flight software is a different world from web services and mobile apps.
> Let's check ourselves a bit and realize developing flight software is a different world from web services and mobile apps.
I agree, it is far more important in avionics to be provably correct and to eliminate as many bugs as possible. It is exactly BECAUSE these aren't web services or mobile apps that C/C++ shouldn't even be a candidate.
You've made exactly zero arguments to support C/C++ here. Just acted smug and condescending to all of us lowly other developers. Do you yourself work in avionics?
I doesn't matter much what language you're using if the HUD blurs during high G forces, clocks run out of sync or you forgot to design a feature or designed the prioritization of events wrongly.