GCC Finishing Up C++17 Adjustments, Preparing For C++2A

[pal666]

C++ standard should not only be a lot more (pro)active at deprecating and removing archaic cruft, but at making and supporting (devs, financially, structure, etc) reliable code upgrading tools to newer standards.

and where does standard get money for this in you land of unicorns and rainbows?

- I also consider fixed year releases should be deprecated, it's part of an archaic thinking. After all they fail all deadlines to complete it as ISO before the year. I consider they should use a Vulkan approach as in release early and release often and accept lots of feedback to make the final versions.


Fixed FTW!

i already asked whether you will be helping to do that in other(still blocked) post. here i will ask other question. what will you do with weekly standard release in practice? like, new week passed, new standard released, what is your next step?

[starshipeleven]

C is simple and elegant.

Bull, it's just simple.

Pointers are one of the best features. When you've written lots of assembly they're easy to understand.

Yeah, because every modern-day programmer has written "lots of assembly" before even starting with C. (Hint: it's not like this).

It's unfortunate that hordes of Java programmers let loose on C give it a bad reputation.

Java programmers don't usually work in areas where C is also a choice. Java programmers are far more likely to use Python/Go/PHP or others in the same boat as that's what companies that must push out complex code very fast use.

[torsionbar28]

What we need, is an entire operating system written in Javascript. You could run it via web browser. It would be so virtualized, you don't even need hardware any more.

[starshipeleven]

What we need, is an entire operating system written in Javascript. You could run it via web browser. It would be so virtualized, you don't even need hardware any more.

Node.js can probably do that. They even have their own OS (userspace on top of Linux kernel). https://node-os.com/

[RealNC]

It does. C is always in the way. Even for shooting yourself in the foot C++ is much efficient at really fucking things up.

Once C grows up, it might get destructors. For now, here's a lollipop.

[audir8]

As someone who left C/C++ behind for Java/Scala soon after college, it's been fun to watch C++ get more functional with C++14, but transactional memory, really? There is a reason why AMD never did an alternative to Intel's TSX instructions. Write immutable code where you can and do your own damn fine-grained locking when you can't, transactional memory is for lazy developers and a bad concurrency paradigm.

I'm all for trying new things in languages, and I wish languages could get more feedback on which features are more appropriately used. I don't think there is about to be a C++EE because of modules and reflection, nor are any Go/Scala/Haskell developers about to jump to C++ because of coroutines, designated initializers or concepts. Frankly a default install including a standard library and a tutorial on smart pointers/boehm would go a long way to attracting developers..

RealNC C doesn't have classes, and doesn't need destructors.

[danieru]

Programmers do not have to use all the language features, just the ones they need or the ones that they fell more comfortable with

A complete lie when it comes to working on open source projects though.

[RealNC]

C doesn't have classes, and doesn't need destructors.

I fail to see how these two statements relate to each other :-/

[audir8]

I fail to see how these two statements relate to each other :-/

I meant C++ style class destructors, but I think you mean try-with-resources or some other resource management nicety. I found this: https://phoxis.org/2011/04/27/c-lang...tors-with-gcc/ and there are ways of doing try-catch-finally in C.

[RealNC]

I meant C++ style class destructors, but I think you mean try-with-resources or some other resource management nicety. I found this: https://phoxis.org/2011/04/27/c-lang...tors-with-gcc/ and there are ways of doing try-catch-finally in C.

Something in the standard itself would be more helpful that hack-ish looking GCC extensions though. And constructors aren't that important for a C-purist. Destructors, however, are very helpful indeed, at least IMO:

if (foo) goto error_cleanup_this;
if (bar) goto error_cleanup_this_other_thing;
if (baz) goto error_cleanup_that_thing_but_not_the_other_stuff;
if (xyzzy) goto climb_on_chair_and_hang_myself;
/* And when refactoring code, some of this will probably break. */

That's just one example where destructors would be quite helpful.

After using C++ for so long, I'm now in a state where I feel like I'm physically incapable of using ANY language that doesn't have destructors :-/ Tying the lifetime of a resource acquisition or state change to the current stack frame feels like the most natural and elegant way to deal with resource cleanup. Not saying it IS the best and most natural way, it just feels that way to me.