Then write an exception handler, and log the exception to a file somewhere.
I'M saying that if your code does do something like that, then it's ALMOST
ABSOLUTELY CERTAINLY going to have a logic error. Even if you prevent the
program from crashing, how can you be sure it's executing properly?
Simplified Example to Establish a Point:
Contact* myContact;
myContact->age = 21;
Is your code executing properly?
IF the OS ignores the access violation, then the Age value will most
certainly be garbage. What if it happened to be a negative number? What if
some code written by another progammer then crashes or corrupts a data file
because that programmer didn't ever expect a negative age?
What do you tell your customer when your program is getting incorrect
results? What if this was a financial application? What if they were
depending on correct results, and assumed that because your program did not
give an error that everything was okay.
With proper exception handling, you can GRACEFULLY handle the access
violation, so the program continues to execute. There is still a problem,
but it won't be a show stopper. Also, you can warn the user that something
may have gone wrong and they MIGHT want to check the results. Better yet,
they can send you a report of the problem so MAYBE you can fix it in Version
2. As a user, I sure as hell prefer to be notified when a program might be
doing something wrong.
What about buffer overruns? If the invalid pointer happened to be pointing
INSIDE your application's allocated memory, it wouldn't generate an Access
violation anyway. If some hacker figured out the bug, they could certainly
exploit your software to execute malicious code.
An access violation is still a sign of something NOT WORKING PROPERLY.
Whether or not your program crashes is up to you as a programmer. It's just
that the Windows XP operating system by default will halt the application.
Linux chooses to log the error. But what end-user would check the log? How
would you ever know your program DID have a bug? I know users aren't beta
testing for you, but how on earth do you fix a problem if you or your
customer don't even know about it?
Just ignoring an error does not make it go away.
--Matthew W. Jackson
Jeff Relf said:
[ I'm reposting this , I use Individual.NET ,
J.A. Bailo on Earthlink saw it , but I never did ? ! ]
Hi Matthew W. Jackson ,
" Sure , none of us want our programs to crash ,
but in this case if it * didn't * crash
how would we ever find out
that the program is doing something wrong ? "
Jesus ****ing Christ ! !
I don't want my paying customers to be my testers ...
Get it ?
If I decide to plop a char *P = 0; *P = 1;
into my code somewhere ,
I'd expect that code to keep running ...
Because , with customers , that's priority One .
