I just rebuilt a core program that talks to my UNIVAC OS2200 CITA and my VAX ACMS from Linux.
I built this first as a 32-bit binary, and there was noise that I had to cut with a substr function.
In the rebuilding with my MinGW 64-bit Windows cross compiler on Oracle Linux, I had to mine function prototypes out of all the code for exactly this reason.
'for those unaware: c89 has a cool "feature" where, if you try to call a function which doesn't exist, rather than erroring out, the function is implicitly declared as a function with unspecified parameters returning int.'
I'm not testing this, because that is agony.
I'm leaving them a working build, and retiring this month.
This code was written by the living and the dead, and my stewardship is finished. I do not envy the poor soul who must test this.
It was ashes when I excavated it myself. A hat tip to those who have gone before.
> I had to mine function prototypes out of all the code for exactly this reason.
The program that does this is called cproto and it was available back on comp.sources.unix as early as 1989 (if you find yourself with a time machine) and these days is available on nearly every Linux distro.
That feature was already present in K&R C, C89 function prototypes were actually taken from the C++ standardisation process back into C, while keeping that behaviour in place.
I'm not sure what the "which will never be fixed" part in the title is about, the problem was clearly fixed in C99 by removing the entire "misfeature" from the standard (e.g. GCC and Clang do error out both in C and C++ compilation mode).
My guess is that the implicit declaration was a leftover from a C predecessor (B?) and kept in C to simplify programming in hybrid B/C environments. Just a guess though.
All the weird stuff I have read about C or C++ doing over the years could be summed up with “Well… don’t do that.”
Like almost without fail. I can’t even think of how you’d end up in a scenario like this because a compiler would complain before letting you use an undefined implicit declaration.
Most people build C in g++ before gcc, as it tends to have a better chance of catching most oddities. Then run it past valgrind to double check if something looks suspect or is slowly leaking. =3
That's not possible for 'standard C' though, only for the separate language that's called "the common subset of C and C++". E.g. it works more or less by accident for some specific C code that predates C99, but even that is built with C++ semantics (which has many subtle and some not-so-subtle differences to C semantics).
Instead of trying to build C in C++ mode, turn all the warnings to 11 (and also include warnings that aren't in -Wall -Wextra, like -Wsign-conversion).
Agreed, the point of divergence is already C89, not C99.
The C++ compatibility with C, is "As Close as Possible to C, but no Closer", meaning supporting C should not break the stronger type checking or the improved type system C++ has over C.
As such, besides C89 original differences, compatibility has only been added for C features that don't clash with C++ approach to the same problem, and while the C++ standard library keeps up with ISO C, it also does so taking into consideration C++ features for the implementation of said library functions.
The 'g++ -x c ...' will parse portable source just fine (compiles code as C, but links against libstdc++), and tends to identify poorly structured code with verbose checks. If people default to gcc file extension inference, than it is likely to remain silent when people do something silly.
Tbh I would be very surprised if `g++ -x c` and `gcc` use different language frontends.
One silly side effect of compiling C in C++ mode is that it forces you to cast from void pointers. E.g. the famous example of:
bla_t* bla = (bla_t*)malloc(sizeof(bla_t));
That extra cast is just a silly and completely pointless C++-ism. It's only the tip of the iceberg though (another difference was zero-initialization with `{}` vs `{0}`, but that's harmonized now in C23 so that `{}` is also valid in C, but as soon as you want to use C99 designated init, the 'harmonization' ends).
One critical missing piece of C compilation mode was usually that the compiler didn't warn when using a wrong enum type, but that's also fixed by now (assuming `-Wall -Wextra`, but that should always be enabled anyway, otherwise you're just flying blind), e.g.:
I’m not going to defend all of the ones C(++) has but stuff like deref’ing null or integer overflow or out of bounds access has its roots in sanity, it just looks insane today. There are tools for many/most of these and if you’re not using them you’re doing it wrong.
I have a 30 page document of errors and ambiguities in the C89 standard -- I joined X3J11 with the intention to present it as a "Monday paper" (papers from committee members submitted on the first meeting day) but I couldn't get a flight and didn't arrive until Tuesday, and they (PJ Plauger specifically) wouldn't accept the paper.
At the next meeting I nevertheless voted to accept the standard. (Being first in alphabetical order, I was the first person to ever do so.)
op implies the issues in c89 were fixed with c99 not c11.
it may seem like a slightly more ridiculous claim, given the infamous features added in c99, although, in hindsight, every version after c89 was less C–like than the previous.
In c89, there doesn't seem to be any distinction between pointer to functions and pointer to data.
Do you know what c89 does explicitly say is undefined? Any pointer arithmetic at all (or array subscripting) for pointers are not pointing to a member of an array. It's already invalid to do pointer arithmetic a pointer to a struct.
Not that it matters for this. This isn't doing pointer arithmetic, f is simply the identifier for the array of "sizeof(f) ints" that make up the function argument. The identifier is optional for a function declaration.
There is a distinction. Lisp machines do not allow function pointers to become data pointers. The C standard accounts for that. Function pointers also have special decay behavior unlike data pointers. You can chain an unlimited number of '*'s and still get a valid function pointer.
int f(int x) { return x + 1; }
int main(int argc, char *argv[]) {
int (*fp)(int) = f;
printf("Answer: %d\n", fp(39) + (****fp)(1));
return 0;
}
Not sure about later versions, but c89 doesn't seem to forbid casting a function pointer to a data pointer (it does say that casting pointers has implementation defined aspects, so lisp machines aren't out of spec if they fail).
The reason you can chain unlimited ''s is that f isn't actually a function pointer, it's a function designator. Function designators are automatically converted to function pointers in almost every case, but '' is special cased so that indirecting a raw function designator results in another function designator.
Edit: Actually on a closer reading, maybe it does. C89 doesn't explicitly point it out as forbidden or undefined, it simply defines two classes (objects and functions) and only defines a very limited set of valid operations on pointers to functions.
Though one operation that is implementation defined is converting a pointer to an int, and an int to a pointer. So with two steps through an implementation defined path, you have a data pointer to a function.
void*
and
void(*)()
are just not guaranteed to be the same size. If the latter is larger then those "implementation defined aspects" are what kicks in. It's nothing to do with the int.
In some systems functions are smaller, since they're gate addresses in a jump table (who has more than 64k functions?), so that's no problem.
But in some systems functions are bigger, and PCDOS systems running in real mode this was very common, and in those cases you'd need to get the extra bits from someplace else (union trick), or just design things another way.
I just rebuilt a core program that talks to my UNIVAC OS2200 CITA and my VAX ACMS from Linux.
I built this first as a 32-bit binary, and there was noise that I had to cut with a substr function.
In the rebuilding with my MinGW 64-bit Windows cross compiler on Oracle Linux, I had to mine function prototypes out of all the code for exactly this reason.
'for those unaware: c89 has a cool "feature" where, if you try to call a function which doesn't exist, rather than erroring out, the function is implicitly declared as a function with unspecified parameters returning int.'
I'm not testing this, because that is agony.
I'm leaving them a working build, and retiring this month.
This code was written by the living and the dead, and my stewardship is finished. I do not envy the poor soul who must test this.
It was ashes when I excavated it myself. A hat tip to those who have gone before.
> I had to mine function prototypes out of all the code for exactly this reason.
The program that does this is called cproto and it was available back on comp.sources.unix as early as 1989 (if you find yourself with a time machine) and these days is available on nearly every Linux distro.
Implicit declarations are a classic!
Current GCC and clang error on them by default now, but that's a rather recent change.
That feature was already present in K&R C, C89 function prototypes were actually taken from the C++ standardisation process back into C, while keeping that behaviour in place.
I'm not sure what the "which will never be fixed" part in the title is about, the problem was clearly fixed in C99 by removing the entire "misfeature" from the standard (e.g. GCC and Clang do error out both in C and C++ compilation mode).
My guess is that the implicit declaration was a leftover from a C predecessor (B?) and kept in C to simplify programming in hybrid B/C environments. Just a guess though.
All the weird stuff I have read about C or C++ doing over the years could be summed up with “Well… don’t do that.”
Like almost without fail. I can’t even think of how you’d end up in a scenario like this because a compiler would complain before letting you use an undefined implicit declaration.
I sum it up with: give me an error message rather than undefined or unintuitive behavior.
That doesn't make it any less interesting :)
It does. Some people would still be interested but lots of people would not. The net interestingness would certainly decrease.
Most people build C in g++ before gcc, as it tends to have a better chance of catching most oddities. Then run it past valgrind to double check if something looks suspect or is slowly leaking. =3
> Most people build C in g++ before gcc
That's not possible for 'standard C' though, only for the separate language that's called "the common subset of C and C++". E.g. it works more or less by accident for some specific C code that predates C99, but even that is built with C++ semantics (which has many subtle and some not-so-subtle differences to C semantics).
Instead of trying to build C in C++ mode, turn all the warnings to 11 (and also include warnings that aren't in -Wall -Wextra, like -Wsign-conversion).
Agreed, the point of divergence is already C89, not C99.
The C++ compatibility with C, is "As Close as Possible to C, but no Closer", meaning supporting C should not break the stronger type checking or the improved type system C++ has over C.
As such, besides C89 original differences, compatibility has only been added for C features that don't clash with C++ approach to the same problem, and while the C++ standard library keeps up with ISO C, it also does so taking into consideration C++ features for the implementation of said library functions.
"Sibling Rivalry: C vs C++"
https://www.stroustrup.com/sibling_rivalry.pdf
"C and C++: Siblings"
https://www.stroustrup.com/siblings_short.pdf
"C and C++, a case for compatibility"
https://www.stroustrup.com/compat_short.pdf
Two examples of the library updates,
"C++17 should refer to C11 instead of C99"
https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2016/p00...
"C++26 should refer to C23 not C17"
https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2025/p33...
The 'g++ -x c ...' will parse portable source just fine (compiles code as C, but links against libstdc++), and tends to identify poorly structured code with verbose checks. If people default to gcc file extension inference, than it is likely to remain silent when people do something silly.
Best of luck, =3
https://www.man7.org/linux/man-pages/man1/g++.1.html
Tbh I would be very surprised if `g++ -x c` and `gcc` use different language frontends.
One silly side effect of compiling C in C++ mode is that it forces you to cast from void pointers. E.g. the famous example of:
That extra cast is just a silly and completely pointless C++-ism. It's only the tip of the iceberg though (another difference was zero-initialization with `{}` vs `{0}`, but that's harmonized now in C23 so that `{}` is also valid in C, but as soon as you want to use C99 designated init, the 'harmonization' ends).One critical missing piece of C compilation mode was usually that the compiler didn't warn when using a wrong enum type, but that's also fixed by now (assuming `-Wall -Wextra`, but that should always be enabled anyway, otherwise you're just flying blind), e.g.:
https://www.godbolt.org/z/h5s4qWoTh
Ray William Johnson…?
Google Gemini? =3
https://www.youtube.com/watch?v=X6WHBO_Qc-Q
What language doesn’t have its own set of absolutely monstrous footguns? https://github.com/denysdovhan/wtfjs - https://github.com/satwikkansal/wtfpython - https://news.ycombinator.com/item?id=31216080 - https://www.dolthub.com/blog/2023-08-16-go-pitfalls/ are all examples of pitfalls of a few languages.
I’m not going to defend all of the ones C(++) has but stuff like deref’ing null or integer overflow or out of bounds access has its roots in sanity, it just looks insane today. There are tools for many/most of these and if you’re not using them you’re doing it wrong.
I have a 30 page document of errors and ambiguities in the C89 standard -- I joined X3J11 with the intention to present it as a "Monday paper" (papers from committee members submitted on the first meeting day) but I couldn't get a flight and didn't arrive until Tuesday, and they (PJ Plauger specifically) wouldn't accept the paper.
At the next meeting I nevertheless voted to accept the standard. (Being first in alphabetical order, I was the first person to ever do so.)
"because the problematic behavior was removed from the standard over 27 years ago, we'll never have closure on what the intended interpretation is."
Obsolete standard turns out to be insufficiently strict, news at 11.
newc at 11, too. it was called c11.
op implies the issues in c89 were fixed with c99 not c11.
it may seem like a slightly more ridiculous claim, given the infamous features added in c99, although, in hindsight, every version after c89 was less C–like than the previous.
Visit the homepage https://sebsite.pw/
That's a nice bit of understated 90s design right there
90s design had frames, tiled backgrounds, low contrast and lots of different font colors.
The "nice" sites didn't style much at all, which usually meant the default serif font was used.
This... from the time when kids played with real toys :)
> f[...]
It is illegal to perform pointer arithmetic on function pointers. They are not necessarily interchangeable with data pointers. Clang is wrong.
In c89, there doesn't seem to be any distinction between pointer to functions and pointer to data.
Do you know what c89 does explicitly say is undefined? Any pointer arithmetic at all (or array subscripting) for pointers are not pointing to a member of an array. It's already invalid to do pointer arithmetic a pointer to a struct.
Not that it matters for this. This isn't doing pointer arithmetic, f is simply the identifier for the array of "sizeof(f) ints" that make up the function argument. The identifier is optional for a function declaration.
There is a distinction. Lisp machines do not allow function pointers to become data pointers. The C standard accounts for that. Function pointers also have special decay behavior unlike data pointers. You can chain an unlimited number of '*'s and still get a valid function pointer.
Not sure about later versions, but c89 doesn't seem to forbid casting a function pointer to a data pointer (it does say that casting pointers has implementation defined aspects, so lisp machines aren't out of spec if they fail).
The reason you can chain unlimited ''s is that f isn't actually a function pointer, it's a function designator. Function designators are automatically converted to function pointers in almost every case, but '' is special cased so that indirecting a raw function designator results in another function designator.
Edit: Actually on a closer reading, maybe it does. C89 doesn't explicitly point it out as forbidden or undefined, it simply defines two classes (objects and functions) and only defines a very limited set of valid operations on pointers to functions.
Though one operation that is implementation defined is converting a pointer to an int, and an int to a pointer. So with two steps through an implementation defined path, you have a data pointer to a function.
void* and void(*)() are just not guaranteed to be the same size. If the latter is larger then those "implementation defined aspects" are what kicks in. It's nothing to do with the int.
In some systems functions are smaller, since they're gate addresses in a jump table (who has more than 64k functions?), so that's no problem.
But in some systems functions are bigger, and PCDOS systems running in real mode this was very common, and in those cases you'd need to get the extra bits from someplace else (union trick), or just design things another way.