[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] [PATCH] introduce VIR_CLOSE to be used rather than close()

On 10/15/2010 11:32 AM, Eric Blake wrote:
+int virClose(int *fdptr);

Needs an ATTRIBUTE_NONNULL(1). I'm also debating whether it needs
ATTRIBUTE_RETURN_CHECK; normally, close() fails in very few situations,
and in cleanup paths, you tend to already have another error more
important so you can ignore the secondary close() failures. So I'm
probably 80-20 against adding ATTRIBUTE_RETURN_CHECK.

+# define VIR_CLOSE(FD) \
+ virClose(&(FD))

Thinking about it a bit more, what if we provide two variants?

int virClose(int *fdptr, bool preserve_errno)
    int saved_errno;
    int rc;

    if (*fdptr >= 0) {
        if (preserve_errno)
            saved_errno = errno;
        rc = close(*fdptr);
        *fdptr = -1;
        if (preserve_errno)
            errno = saved_errno;
    } else
        rc = 0;

    return rc;

int virClose(int *fdptr) ATTRIBUTE_RETURN_CHECK;

/* For use on normal paths; caller must check return value, and failure sets errno per close(). */
#define VIR_CLOSE(FD) virClose(&(FD),false);

/* For use on cleanup paths; errno is unaffected by close, and no return value to worry about. */
#define VIR_FORCE_CLOSE(FD) ignore_value(virClose(&(FD),true))

Eric Blake   eblake redhat com    +1-801-349-2682
Libvirt virtualization library http://libvirt.org

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]