Giving C a superpower: custom header file (safe_c.h)

Nice toy. It works until it stops working. An experienced C developer would quickly find a bunch of corner cases where this just doesn't work.

Given how simple examples in this blog post are, I ask myself, why don't we already have something like that as a part of the standard instead of a bunch of one-off personal, bug-ridden implementations?

Yeah, kids like to waste time to make C more safe or bring C++ features. If you need them, use C++ or different language. Those examples make code look ugly and you are right, the corner cases.

If you need to cleanup stuff on early return paths, use goto.. Its nothing wrong with it, jump to end when you do all the cleanup and return. Temporary buffers? if they arent big, dont be afraid to use static char buf[64]; No need to waste time for malloc() and free. They are big? preallocate early and reallocate or work on chunk sizes. Simple and effective.

It would be a lot more constructive if you reported a bunch of corner cases where this doesn't work rather than just dismissing this as a toy.

Can you share such a corner case?

God forbid we should make it easier to maintain the existing enormous C code base we’re saddled with, or give devs new optional ways to avoid specific footguns.

Goofy platform specific cleanup and smart pointer macros published in a brand new library would almost certainly not fly in almost any "existing enormous C code base". Also the industry has had a "new optional ways to avoid specific footguns" for decades, it's called using a memory safe language with a C ffi.

I meant the collective bulk of legacy C code running the world that we can’t just rewrite in Rust in a finite and reasonable amount of time (however much I’d be all on board with that if we could).

There are a million internal C apps that have to be tended and maintained, and I’m glad to see people giving those devs options. Yeah, I wish we (collectively) could just switch to something else. Until then, yay for easier upgrade alternatives!

No, I don't dismiss anything.

It's just, I'd rather play with my own toys instead of using someone else's toy. Especially since I don't think it would ever grow up to be something more than a toy.

For serious work, I'd use some widely used, well-maintained, and battle-tested library instead of my or someone else's toy.

I was also, in fact, referring to the bulk of legacy code bases that can't just be fully rewritten. Almost all good engineering is done incrementally, including the adoption of something like safe_c.h (I can hardly fathom the insanity of trying to migrate a million LOC+ of C to that library in a single go). I'm arguing that engineering effort would be better spent refactoring and rewriting the application in a fully safe language one small piece at a time.

> static char buf[64];

In a function? That makes the function not-threadsafe and the function itself stateful. There are places, where you want this, but I would refrain from doing that in the general case.

I’m not sure I agree with that, especially if there were easy wins that could make the world less fragile with a much smaller intermediate effort, eg with something like FilC.

I wholeheartedly agree that a future of not-C is a much better long term goal than one of improved-C.

> use goto

My thoughts as well. The only thing I would be willing to use is the macro definition for __attribute__, but that is trivial. I use C, because I want manual memory handling, if I wouldn't want that I would use another language. And now I don't make copies when I want to have read access to some things, that is simply not at a problem. You simply pass non-owning pointers around.

I don't really agree, at least if the future looks like Rust. I much prefer C and I think an improved C can be memory safe even without GC.

> I think an improved C can be memory safe even without GC

That's a very interesting belief. Do you see a way to achieve temporal memory safety without a GC, and I assume also without lifetimes?

A simple pointer ownership model can achieve temporal memory safety, but I think to be convenient to use we may need lifetimes. I see no reason this could not be added to C.

A C with lifetimes would be nice, I agree.

Would be awesome if someone did a study to see if it's actually achievable... Cyclone's approach was certainly not enough, and I think some sort of generics or a Hindley-Milner type system might be required to get it to work, otherwise lifetimes would become completely unusable.

No, because I did NOT do serious analisis of this. Nor I care, ask upper commenter.. C have some corner case and undefined behaviours and this stuff will make it worse IMO.

Holy moly.. Thread safety.. Good point and Bad point. I myself use threads sparsly, so I dont intermix calls between threads..

Yes, one needs polymorphism. Let's see. I have some ideas.

It also has different behaviour in a single thread. This can be what you want though, but I would prefer it to pass that context as a parameter instead of having it in a hidden static variable.

C does have the concept of lifetimes. There is just no syntax to specify it, so it is generally described along all the other semantic details of the API. And no it is not the same as for Rust, which causes clashes with the Rust people.

What different behaviour you mean? static in function means that this is just preallocated somewhere in data, not on stack nor heap. Thats it. Yes, its not thread safe so should never be used in libraries for any buffering.

But in program, its not bad. If I ever need multiple calls to it in same thread:

  static char buf[8][32];
  static int z;
  char *p=buf[z];
  z=(z+1)&7;

Works pretty well ;)

> What different behaviour you mean?

Static foremost means that the value is preserved from the last function invocation. This is very different behaviour, than an automatically allocated variable. So calling a function with a static variable isn't idempotent, even when all global variables are the same.

> If I ever need multiple calls to it in same thread:

What is this code supposed to do???? It hands out a different pointer, the first 8 times, than starts from the beginning again? I don't see what this is useful for!

What clashes?

I think there was a discussion in the Linux kernel between a kernel maintainer and the Rust people, which started by the Rust people demanding formal semantics, so that they could encode it in Rust, and the subsystem maintainer unwilling to do that.

If you want to return some precalculated stuff w/o using malloc() free(). So you just have 8 preallocated buffers and you rotate them between calls. Of course you need to be aware that results have short lifetime.

Ah, I thought you were talking about core language semantics in C.

I don't know enough Rust, to do such a comparison.

That sounds like a maintenance nightmare to me. If you insist on static, I would at least only use one buffer, to make it predictable, but personally I would just let the caller pass a pointer, where I can put the data.

What application is that for? Embedded, GUI program, server, ...?

“The rust people” were also kernel maintainers.

Sorry, I'm not familiar with the titles of kernel developers. I thought only one of them was the subsystem maintainer.

One of them was a maintainer of that particular subsystem, but that doesn't mean that the other folks aren't also maintainers of other parts of the kernel.

it is very predicatable.. Every call it returns buffer, after 8 calls it wraps. I use such stuff in many places.. GUI programs.. daemons.. Most stuff are single threaded. If threads are used, they are really decupled from each other.

Yes, you should never ever use it in Library.. But small utility functions should be okish :)

This is example from my Ruby graph library. GetMouseEvent can be called alot, but I need at most 2 results. Its Ruby, so I can either dynamicaly allocate objects and let GC pickup them later, or just use static stuff here, no GC overhead. it can be called 100s of times per second, so its worth it.

  static GrMouseEvent evs[8];
  static int z=0;
  GrMouseEvent *ev=&evs[z];