svn commit: r303586 - head/bin/sh

[Bruce Evans]

On Sun, 31 Jul 2016, Warner Losh wrote:

> On Sun, Jul 31, 2016 at 2:16 PM, Jilles Tjoelker <jilles at stack.nl> wrote:
>> On Sun, Jul 31, 2016 at 01:43:16PM +0000, Alexey Dokuchaev wrote:
>>> On Sun, Jul 31, 2016 at 01:11:34PM +0000, Jilles Tjoelker wrote:
>>>> New Revision: 303586
>>>> URL: https://svnweb.freebsd.org/changeset/base/303586
>>
>>>> Log:
>>>>   sh: Fix a clang warning.
>>
>>>>   Submitted by:     bdrewery
>>
>>>> Modified:
>>>>   head/bin/sh/expand.c
>>
>>>> Modified: head/bin/sh/expand.c
>>>> ==============================================================================
>>>> --- head/bin/sh/expand.c    Sun Jul 31 12:59:10 2016        (r303585)
>>>> +++ head/bin/sh/expand.c    Sun Jul 31 13:11:34 2016        (r303586)
>>>> @@ -473,7 +473,8 @@ expbackq(union node *cmd, int quoted, in
>>>>             if (--in.nleft < 0) {
>>>>                     if (in.fd < 0)
>>>>                             break;
>>>> -                   while ((i = read(in.fd, buf, sizeof buf)) < 0 && errno == EINTR);
>>>> +                   while ((i = read(in.fd, buf, sizeof buf)) < 0 && errno == EINTR)
>>>> +                           ;
>>
>>> `continue;' would be even better; some tools might barf at stray semicolon.
>>
>> Both continue; and ; (the latter with and without comment) occur in the
>> source tree many times. I don't really like a continue that does nothing
>> because it is at the end of a loop, so I prefer to make this whitespace
>> change only (there are two more instances in bin/sh). I think a sole
>> semicolon on a line is conspicuous enough that nothing additional is
>> required.
>
> For humans, yes. For picky tools that warn about strange constructs, no.
> Clang may be happy, but there's many other tools that expect you to declare
> an 'empty' while loop with continue. This tradition has dated back to
> at least the
> late 80's...

Buggy tools.  I thought that programmers used the stand-alone semicolon
since that is shorter and clearer.

The stand-alone semicolon is bad if it is misformatted.  'while(foo());'
is shorter and unclearer.  A C parser must ignore whitespace, so you
need a tool like indent that sort of understands whitespace to disallow
while(foo()); but accept 'while (foo())\n\t;'.  It is not far from full
indent(1) and insisting on the correct number of \t's before the semicolon.

Bruce