On 2018-09-12 at 16:25, Shi Lei wrote: On 2018-09-12 at 15:21, Erik Skultety wrote:

On Tue, Sep 11, 2018 at 11:37:45PM +0800, Shi Lei wrote:

...

On 2018-09-11 at 20:44, Erik Skultety wrote:

...

On Mon, Sep 10, 2018 at 11:47:53AM +0800, Shi Lei wrote:

...

By making use of GNU C's cleanup attribute handled by the VIR_AUTOCLOSE macro, many of the VIR_FORCE_CLOSE calls can be dropped, which in turn leads to getting rid of many of our cleanup sections.

Signed-off-by: Shi Lei <shi_lei@massclouds.com> ---  static int @@ -909,9 +867,9 @@ char *virNetDevGetName(int ifindex)   #if defined(SIOCGIFINDEX) && defined(HAVE_STRUCT_IFREQ)   int virNetDevGetIndex(const char *ifname, int *ifindex)   { -    int ret = -1;       struct ifreq ifreq; -    int fd = socket(VIR_NETDEV_FAMILY, SOCK_DGRAM, 0); +    VIR_AUTOCLOSE(fd); +    socket(VIR_NETDEV_FAMILY, SOCK_DGRAM, 0);

^this could not potentially work...

Erik

Sorry! I'm too careless.

I think that a new syntax-check rule might make sense. This rule checks below:

type foo(a0, a1 ...);

[type] var = foo(a0, a1, ...);  /* It's OK */ ignore_value(foo(a0, a1, ...)); /* It's OK */

foo(a0, a1, ...); /* Report this usage */

but this would go away with the syntax-check rule you proposed in you response to the first patch? If not, then would you be more specific to help me understand the problem more?

Thanks, Erik

Sorry! I was muddle-headed last night. It is not the syntax-check rule in my response to the first patch. This is another rule and I wanted to introduce it to avoid some faults then. Today I find that it is not easy as I thought. Please ignore it ...  :-) Shi Lei

...

And it takes effect before compilation and even it's in the inactive side of the #if-#else conditons.

It would not check macros since their name are upper case and we don't care for the function as condition in if-else statement.

How about it? And I can try it if it's a bit helpful ...

Shi Lei

-- libvir-list mailing list libvir-list@redhat.com https://www.redhat.com/mailman/listinfo/libvir-list