Re: Realloc is blocking execution

[ERSEK Laszlo <lacos@caesar.elte.hu>]

On Tue, 13 Oct 2009, Felipe Sateler wrote:

> I've always assumed that malloc (and thus realloc) can sleep.

I believe the creators of the Single UNIX Specification didn't share (or 
at least, didn't codify) this assumption. As a general rule, any system 
interface that is designed to wait (= hang indefinitely) on a resource

- can be either used in some non-blocking way when it returns EAGAIN,
- or it can be interrupted by a signal, when it returns -1 and sets errno
  to EINTR,
- or you can specify a time limit,
- or it is a thread cancellation point.

In short, if a system interface goes to sleep by design, it can be woken 
up in a way or another, or else the system interface description 
explicitly says it cannot be woken up. The malloc() description under [0] 
makes me think that malloc() cannot hang indefinitely by design.

[0] http://www.opengroup.org/onlinepubs/9699919799/functions/malloc.html

Cheers,
lacos