svn commit: r185356 - head/sys/dev/ixgbe

[Bruce Evans]

On Thu, 27 Nov 2008, [utf-8] Dag-Erling Smørgrav wrote:

> "M. Warner Losh" <imp at bsdimp.com> writes:
>> I personally really dislike the style (and yes, I know all the
>> arguments for it).  If you really want something that complex inside a
>> block to need block scoped variables, then that really argues for a
>> function oft times...
>
> There is one particular situation where it is very convenient:
>
> int
> foo(struct sockaddr *sa)
> {
>        switch (s->sa_family) {
>        case AF_INET: {
>                struct sockaddr_in *sin = (struct sockaddr_in *)sa;
>                /* ... */
>                break;
>        }

Switch statements are 1 case (the only case?) where the braces needed for
non-C99 declarations don't require extra indentation.  However, indent(1)
doesn't understand this and mangles the above to:

int
foo(struct sockaddr *sa)
{
 	switch (s->sa_family) {
 		case AF_INET:{
 			struct sockaddr_in *sin = (struct sockaddr_in *)sa;

 			/* ... */
 			break;
 		}

indent(1) has fixed a style bug in the above (the missing blank line after
the declarations) but it has introduced 2 (the missing space before "{"
and the unwanted indentation).  Fixing the other style bug in the above
(the initialization in the declaration) is beyond the scope of indent(1).
gindent-2.2.9 -npro gives the same style bugs and 1 more (it misformats
the comment to a block one, but that may be due to a bad default).

Bruce