Thread: how can I write code to test a[100];

Originally Posted by zcrself

Sometimes, when my program runs, there is a error "segmentation error", but I don't know why and where is the error. I want to handle this error, but I don't know how to do handling.

That indicates a programming error that should be fixed. As nimitzhunter noted, a debugger is suited to the task.

Originally Posted by zcrself

java language has try-catch function,
what does C language has?

In standard C, accessing an array out of bounds results in undefined behaviour. There is no such exception handling feature. In the event that you use Java instead, you should still fix such programming errors rather than rely on them being handled somewhere by some exception handler.

For what it's worth, you could try this.

Code:

#include <stdio.h>
#include <signal.h>
#include <setjmp.h>

sigjmp_buf env;

void trap ( int sig )
{
  write(1,"Caught!\n",8);
  siglongjmp(env,1);
}

int main()
{
  struct  sigaction sa, oldsa;
  sa.sa_handler = trap;
  sigemptyset(&sa.sa_mask);
  sa.sa_flags = 0;
  int r = sigaction(SIGSEGV,&sa,&oldsa);
  if ( sigsetjmp(env,1) == 0 ) {
    int *p = NULL;
    *p = 0;
  } else {
    printf("Avoided the NULL dereference\n");
  }
  return 0;
}


$ ./a.out 
Caught!
Avoided the NULL dereference

But your program was screwed when it made it's first out of bound memory access (but still within a valid page).

Since it's been mentioned so many times already, let me repeat it.

You need to DEBUG your code so it doesn't do these things. Not attempt some half-assed recovery mechanism when lots of damage has already been done. What is certain, there will be NO way to recover the damage done between the end of the array (you walked off the end of), and the end of the page it resides within.

And it could be a hell of a lot of damage as well. Pages are allocated sequentially (most of the time), so you're only going to get the SIGSEGV when you walk off the end of the last allocated page.

Originally Posted by Adak

Smarter programmers!

Ooh...nice burn!

Originally Posted by zcrself

java language has try-catch function,
what does C language has?

C has undefined behaviour

Seriously, Java (more accurately, the Java virtual machine, not the language) imposes a runtime overhead on all programs, in order to detect buffer overruns.

C is not associated with any virtual machine, so a C program is allowed to run over memory, without being forced to stay within array bounds. A segmentation fault (under unix) means the operating system has detected a program accessing some memory that it shouldn't. If the operating system does not detect such things (which is possible, as operating systems do not typically provide 100% coverage either of such things) they will remain undetected.

The trade-off is one of performance versus need for programmer discipline. Java trades off performance in order to capture and report programmer errors. C relies on a programmer to do things right, but - if the programmer does things right - does not impose significant performance overhead.

Note: despite marketing spiels suggesting Java is safe, it is actually technically possible for a buffer overrun to remain undetected in Java (to get a 100% guarantee, it would be necessary to sacrifice a lot more in terms of performance and other attributes). The only difference is that an overrun is more likely to be detected at the point it happens in Java, when compared with C.

Originally Posted by zcrself

Can "try-catch" function handle the error "segmentation fault"

In C++, it cannot. In C, this is a meaningless question.

Accessing data past the end of an array is undefined behaviour, in both C and C++. Once you're in the realms of undefined behaviour, there is no guarantee than any feature of the language is even usable. That includes exceptions: there is no means defined in the standard by which an invalid pointer operation causes any exception to be thrown.

As noted above, the onus is on the C and C++ programmer to prevent or avoid instances of undefined behaviour, rather than relying on detection of such circumstances. For this reason, there is no standard means of detecting things like buffer overruns.

Originally Posted by Adak

Smarter programmers!

that is funny.

Thanks a lot for a lot of senior programmers answering my questions eagerly.
I learn C further for a newbie of C