Reducing the size of objects with realloc()

[Spiros Bousbouras]

If you do realloc(p , a) where a is smaller than the size
of the object p points to is there any realistic possibility
that realloc will return 0 ? Is it worth checking for ?

 

[Harald van Dijk]

If you do realloc(p , a) where a is smaller than the size of the object
p points to is there any realistic possibility that realloc will return
0 ? Is it worth checking for ?

Possibly, and yes. If blocks of smaller sizes are allocated in different
regions of memory, and you have lots of small blocks already, then
realloc can fail when shrinking objects. It's a small effort to check for
0 and use the old value, if that's good enough for you.

 

[qarnos]

If you do realloc(p , a) where a is smaller than the size
of the object p points to is there any realistic possibility
that realloc will return 0 ? Is it worth checking for ?

Is it really worth *not* checking for? What if realloc() decides to
move your object to optimize memory usage?

 

[qarnos]

Is it really worth *not* checking for? What if realloc() decides to
move your object to optimize memory usage?

Scratch that - I misread the question. But I'd still check for NULL.

 

[CBFalconer]

If you do realloc(p , a) where a is smaller than the size of
the object p points to is there any realistic possibility that
realloc will return 0? Is it worth checking for?

Yes, it is possible. The system builder was foolish, but still met
the C standard.