CODEWITHSUNDEEP
capierrno
Eduardo
Eduardo

Reputation: 2341

Do you recommend using errno in new C APIs?

Do you recommend using errno in new C APIs?

Should other methods of reporting problems like http://developer.gnome.org/glib/stable/glib-Error-Reporting.html be used instead?

What do you recommend?

Upvotes: 4

Views: 614

Answers (2)

datenwolf
datenwolf

Reputation: 162319

errno should be reserved to the C standard library, a program should treat it as a read only value.

Writing new code I strongly discourage using a errno-style global variable for error code checking. Say you have functions a, b reporting their status through ab_error then you must write things like this:

a();
if( ab_error == AB_NOERROR ) {
    b();
    if( ab_error == NO_ERROR ) {
        /* ... */
    }
}

You see, that you rather quickly pile up nesting levels. If however your functions returned a error code you can write it this way:

int error_a;
int error_b;

if( (error_a = a()) == A_NOERROR
 && (error_b = b()) == B_NOERROR
){
    /* ... */
} else {
    /* test the error variables */
}

you can even chain multiple error conditions like this:

int error_a;
int error_b;
int error_c;

if( (error_a = a()) == A_NOERROR
 &&  ( (error_b = b()) == B_NOERROR
    || (error_c = c()) == C_NOERROR )
){
    /* ... */
} else {
    /* test the error variables */
}

Which is a lot more readable than if you'd scatter the whole thing over several nested if statements.

Upvotes: 4

Schedler
Schedler

Reputation: 1423

You can always ask yourself how well a global errno works in programs with multiple threads (or other forms of concurrency not containing separate copies of global variables). The typical conclusion is likely to be "not very well", and hence other solutions might be preferable.

Upvotes: 1

Related Questions