Difference between API and ABI

Question

I am new to linux system programming and I came across API and ABI while reading Linux System Programming.

Definition of API :

Answer 1

Your program (source code) can be compiled with modules who provide proper API.

Your program (binary) can run on platforms who provide proper ABI.

API restricts type definitions, function definitions, macros, sometimes global variables a library should expose.

ABI restricts what a "platform" should provide for you program to run on. I like to consider it in 3 levels:

• processor level - the instruction set, the calling convention

• kernel level - the system call convention, the special file path convention (e.g. the /proc and /sys files in Linux), etc.

• OS level - the object format, the runtime libraries, etc.

Consider a cross-compiler named arm-linux-gnueabi-gcc. "arm" indicates the processor architecture, "linux" indicates the kernel, "gnu" indicates its target programs use GNU's libc as runtime library, different from arm-linux-androideabi-gcc which use Android's libc implementation.

Answer 2

Let me give a specific example how ABI and API differ in Java.

An ABI incompatible change is if I change a method A#m() from taking a String as an argument to String... argument. This is not ABI compatible because you have to recompile code that is calling that, but it is API compatible as you can resolve it by recompiling without any code changes in the caller.

Here is the example spelled out. I have my Java library with class A

// Version 1.0.0
public class A {
    public void m(String string) {
        System.out.println(string);
    }
}

And I have a class that uses this library

public class Main {
    public static void main(String[] args) {
        (new A()).m("string");
    }
}

Now, the library author compiled their class A, I compiled my class Main and it is all working nicely. Imagine a new version of A comes

// Version 2.0.0
public class A {
    public void m(String... string) {
        System.out.println(string[0]);
    }
}

If I just take the new compiled class A and drop it together with the previously compiled class Main, I get an exception on attempt to invoke the method

Exception in thread "main" java.lang.NoSuchMethodError: A.m(Ljava/lang/String;)V
        at Main.main(Main.java:5)

If I recompile Main, this is fixed and all is working again.

Answer 3

Linux shared library minimal runnable API vs ABI example

This answer has been extracted from my other answer: What is an application binary interface (ABI)? but I felt that it directly answers this one as well, and that the questions are not duplicates.

In the context of shared libraries, the most important implication of "having a stable ABI" is that you don't need to recompile your programs after the library changes.

As we will see in the example below, it is possible to modify the ABI, breaking programs, even though the API is unchanged.

main.c

#include <assert.h>
#include <stdlib.h>

#include "mylib.h"

int main(void) {
    mylib_mystrict *myobject = mylib_init(1);
    assert(myobject->old_field == 1);
    free(myobject);
    return EXIT_SUCCESS;
}

mylib.c

#include <stdlib.h>

#include "mylib.h"

mylib_mystruct* mylib_init(int old_field) {
    mylib_mystruct *myobject;
    myobject = malloc(sizeof(mylib_mystruct));
    myobject->old_field = old_field;
    return myobject;
}

mylib.h

#ifndef MYLIB_H
#define MYLIB_H

typedef struct {
    int old_field;
} mylib_mystruct;

mylib_mystruct* mylib_init(int old_field);

#endif

Compiles and runs fine with:

cc='gcc -pedantic-errors -std=c89 -Wall -Wextra'
$cc -fPIC -c -o mylib.o mylib.c
$cc -L . -shared -o libmylib.so mylib.o
$cc -L . -o main.out main.c -lmylib
LD_LIBRARY_PATH=. ./main.out

Now, suppose that for v2 of the library, we want to add a new field to mylib_mystrict called new_field.

If we added the field before old_field as in:

typedef struct {
    int new_field;
    int old_field;
} mylib_mystruct;

and rebuilt the library but not main.out, then the assert fails!

This is because the line:

myobject->old_field == 1

had generated assembly that is trying to access the very first int of the struct, which is now new_field instead of the expected old_field.

Therefore this change broke the ABI.

If, however, we add new_field after old_field:

typedef struct {
    int old_field;
    int new_field;
} mylib_mystruct;

then the old generated assembly still accesses the first int of the struct, and the program still works, because we kept the ABI stable.

Here is a fully automated version of this example on GitHub.

Another way to keep this ABI stable would have been to treat mylib_mystruct as an opaque struct, and only access its fields through method helpers. This makes it easier to keep the ABI stable, but would incur a performance overhead as we'd do more function calls.

API vs ABI

In the previous example, it is interesting to note that adding the new_field before old_field, only broke the ABI, but not the API.

What this means, is that if we had recompiled our main.c program against the library, it would have worked regardless.

We would also have broken the API however if we had changed for example the function signature:

mylib_mystruct* mylib_init(int old_field, int new_field);

since in that case, main.c would stop compiling altogether.

Semantic API vs programming API vs ABI

We can also classify API changes in a third type: semantic changes.

For example, if we had modified

myobject->old_field = old_field;

to:

myobject->old_field = old_field + 1;

then this would have broken neither API nor ABI, but main.c would still break!

This is because we changed the "human description" of what the function is supposed to do rather than a programmatically noticeable aspect.

I just had the philosophical insight that formal verification of software in a sense moves more of the "semantic API" into a more "programmatically verifiable API".

Semantic API vs Programming API

We can also classify API changes in a third type: semantic changes.

The semantic API, is usually a natural language description of what the API is supposed to do, usually included in the API documentation.

It is therefore possible to break the semantic API without breaking the program build itself.

For example, if we had modified

myobject->old_field = old_field;

to:

myobject->old_field = old_field + 1;

then this would have broken neither programming API, nor ABI, but main.c the semantic API would break.

There are two ways to programmatically check the contract API:

• test a bunch of corner cases. Easy to do, but you might always miss one.
• formal verification. Harder to do, but produces mathematical proof of correctness, essentially unifying documentation and tests into a "human" / machine verifiable manner! As long as there isn't a bug in your formal description of course ;-)

Tested in Ubuntu 18.10, GCC 8.2.0.