cannot save main .cpp in codeblocks compiler

But that escape hatch should be an opt-in feature, not opt out.

I absolutely understand the need for an escape hatch, be it for performance reasons (at compile time or run time in some hot loop or similar), working around a bug in the checker, or some esoteric reasons. Until safety is explicitly opt out, this is a whole nothing burger. Putting opt-in safety into the language is just checking a box so that they can push back against the NSA with a 'well, actually' and the status quo will not change.Ĭouldn't agree more. There were some efforts in Clang to recover type sugar, but it is a partial solution, it took a lot of work to get there, and it introduced significant compile time and memory usage regressions for template heavy code.Īll in all, I have the feeling most of the safety related proposals are coming from people who did not implement the rules they propose, and I'd prefer to have more discussion with the implementors before taking any serious action. This is a problem both in Clang and in MSVC, (and would not be surprised if GCC had similar problems). As mandated by the standard, std::vector and std::vector> are the same specializations. Moreover, some of the approaches Bjarne is pushing for simply cannot be implemented efficiently in some of the major compilers today. Many of those rules only applicable to narrow circumstances and will break down pretty quickly if people try to apply them more broadly. Large portion of the core guidelines rules have lacking adoption and for a reason. Is it about some of the concrete solutions like gsl::owner? I am strongly against that. In fact, this is how Rust works, you can use safe borrows and unsafe pointers in the same function, and you still have borrow checking in unsafe code).

Is it about how we delineate safe code and unsafe code? It starts to get a bit muddier I don't see many different approaches compared, and there are other alternatives (e.g., instead of marking a region of code safe, you could mark certain types safe, so you can relatively naturally mix safe and unsafe constructs in the same code. Is it about combining static checks, dynamic checks, and libraries? Absolutely, I am on board. It was not clear to me what do we mean by the 'approach'. > Is the approach presented in the paper worth exploring? To me, the question was ambiguous, paraphrasing:

Cannot save main .cpp in codeblocks compiler