3 ms·
Agreed - I do not think the issue is with being technically correct, but in thinking that it always settles the matter and ends the discussion. It would seem p
by mannykannot 10y ago
Agreed - I do not think the issue is with being technically correct, but in thinking that it always settles the matter and ends the discussion.
It would seem particularly unfortunate if a compiler were to do static analysis for undefined behavior, but only use it in optimization. From a cursory search, I found this blog post [1] explaining why LLVM does not warn about such things (at least in 2011, when it was written - attitudes may have changed, and the Clang project now has the UndefinedBehaviorSanitizer [2]). One of the arguments is that it is difficult to explain what the problem is, but in my view, that is an argument for making doing so a priority.
[1] http://blog.llvm.org/2011/05/what-every-c-programmer-should-know_21.html http://blog.llvm.org/2011/05/what-every-c-programmer-should-...
[2] http://releases.llvm.org/3.8.0/tools/clang/docs/UndefinedBehaviorSanitizer.html http://releases.llvm.org/3.8.0/tools/clang/docs/UndefinedBeh...
- kutkloon7 10y agoIMHO, more work should be done on proving properties of programs. If one could make a programming language like Dafny [1] less quirky but more developer-friendly, you can have programs which are both good-performing and proven to be consistent with their specification (but again, quirks in the specification may cause troubles).
- mannykannot 10y agoThanks for bringing Dafny to my attention. I agree with your first sentence, which is why I think it is a big deal that it can be a great deal harder, in practice, to reason about the behavior of a C program, if optimized, than the same program un-optimized. For one thing, unless you successfully reason about what the optimized code does when deploying an optimized build, then you are not deploying the program that you analyzed or inspected. I think there is a sort of analogy with a large black hole here [1]: you can slip across the "event horizon" into undefined behavior without noticing, but, especially if you are using an optimizing compiler, there may be no escape. [1] Maybe not, if the black-hole firewall hypothesis is correct.
- bluejekyll 10y agoI'm totally unfamiliar with Dafny, the front page didn't do much more than specicify the goals. Any chance you could briefly compare it to Rust or another systems level language trying to remove undefined behavior?
- nickpsecurity 10y agoIt's more like SPARK Ada. It's a limited language designed for easy analysis by an automated prover. You can do preconditions, postconditions, and invariants in it. It was used in ExpressOS for mobile and I think IronClad Apps, too.
- bluejekyll 10y agoAre those pre and post conditions always executed? Or are they optional assertions that can be enabled/disabled?
- neilparikh 10y agoIf I'm reading the description correctly, the pre and post conditions are analyzed statically during compiling, and not added to the output code. So it's a similar idea to type checking, where the program won't even compile until the versifier is sure your code meets the conditions.