Kernel generally follow the style
if (func()) {
/* failed case */
} else {
/* success */
}
-aneesh
On 08/10/06, Aneesh Kumar K.V <[email protected]> wrote:
> Kernel generally follow the style
>
> if (func()) {
> /* failed case */
> } else {
> /* success */
> }
>
>
Please submit patches inline, having to copy them from attachments to
be able to reply is a pain.
>
> diff --git a/kernel/sys.c b/kernel/sys.c
> index 98489d8..55cb77c 100644
> --- a/kernel/sys.c
> +++ b/kernel/sys.c
> @@ -517,7 +517,7 @@ EXPORT_SYMBOL_GPL(srcu_notifier_call_cha
> void srcu_init_notifier_head(struct srcu_notifier_head *nh)
> {
> mutex_init(&nh->mutex);
> - if (init_srcu_struct(&nh->srcu) < 0)
> + if (init_srcu_struct(&nh->srcu))
> BUG();
> nh->head = NULL;
> }
I really liked the old code better. If in the future
init_srcu_struct() is changed to also return >0 for some conditions,
then that would not previously have triggered BUG(), but after your
changes it will. The code, as it were, perfectly expressed what it
wanted to happen - if it returns less than zero it's a BUG().
I say leave it alone.
--
Jesper Juhl <[email protected]>
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please http://www.expita.com/nomime.html
>> Kernel generally follow the style
>>
>> if (func()) {
>> /* failed case */
>> } else {
>> /* success */
>> }
Here's my: NAK.
(At best it should be if(foo != 0) rather than if(foo), but that's just me.)
> I really liked the old code better. If in the future
> init_srcu_struct() is changed to also return >0 for some conditions,
> then that would not previously have triggered BUG(), but after your
> changes it will. The code, as it were, perfectly expressed what it
> wanted to happen - if it returns less than zero it's a BUG().
> I say leave it alone.
I agree here.
-`J'
--
Jesper Juhl wrote:
> On 08/10/06, Aneesh Kumar K.V <[email protected]> wrote:
>> Kernel generally follow the style
>>
>> if (func()) {
>> /* failed case */
>> } else {
>> /* success */
>> }
>>
>>
>
> Please submit patches inline, having to copy them from attachments to
> be able to reply is a pain.
>
>>
>> diff --git a/kernel/sys.c b/kernel/sys.c
>> index 98489d8..55cb77c 100644
>> --- a/kernel/sys.c
>> +++ b/kernel/sys.c
>> @@ -517,7 +517,7 @@ EXPORT_SYMBOL_GPL(srcu_notifier_call_cha
>> void srcu_init_notifier_head(struct srcu_notifier_head *nh)
>> {
>> mutex_init(&nh->mutex);
>> - if (init_srcu_struct(&nh->srcu) < 0)
>> + if (init_srcu_struct(&nh->srcu))
>> BUG();
>> nh->head = NULL;
>> }
>
> I really liked the old code better. If in the future
> init_srcu_struct() is changed to also return >0 for some conditions,
> then that would not previously have triggered BUG(), but after your
> changes it will. The code, as it were, perfectly expressed what it
> wanted to happen - if it returns less than zero it's a BUG().
> I say leave it alone.
>
>
As per Documentation/CodingStyle
"Functions can return values of many different kinds, and one of the
most common is a value indicating whether the function succeeded or
failed. Such a value can be represented as an error-code integer
(-Exxx = failure, 0 = success) or a "succeeded" boolean (0 = failure,
non-zero = success)."
That means if the function need to indicate success it should be made to return 0.
I don't see any other value being returned from init_srcu_struct. Also having a consistent
style of if() check make code reading easier.
-aneesh
Aneesh Kumar K.V wrote:
> As per Documentation/CodingStyle
>
> "Functions can return values of many different kinds, and one of the
> most common is a value indicating whether the function succeeded or
> failed. Such a value can be represented as an error-code integer
> (-Exxx = failure, 0 = success) or a "succeeded" boolean (0 = failure,
> non-zero = success)."
>
> That means if the function need to indicate success it should be made
> to return 0.
The wording is 'can', not 'should' or 'shall'. The current agreement is
however that do_something()-named functions indeed 'should' return <0
for "failure" and 0 for "success" while is_something()-named functions
'should' return non-0 for "yes" and 0 for "no". But there is no rule
without exceptions: Some functions like copy_to_user() return >0 in
situations which can be considered a "failure" because this positive
value has further meaning.
But back to your patch: I am not aware of an agreement on how to write a
check for zero or a check for nonzero.
> I don't see any other value being returned from init_srcu_struct.
True.
> Also having a consistent style of if() check make code reading easier.
Also true. However (a) there is no kernel-wide consistency about this
and (b) the style used in the file which you are patching is
if (do_something_returning_negative_errors() < 0)
E.g.
http://www.linux-m32r.org/lxr/http/source/kernel/sys.c?v=2.6.19-rc1#L1070
And here <0 and !0 are actually different:
http://www.linux-m32r.org/lxr/http/source/kernel/sys.c?v=2.6.19-rc1#L852
However, kernel/sys.c is not entirely consequent:
http://www.linux-m32r.org/lxr/http/source/kernel/sys.c?v=2.6.19-rc1#L1336
and the other calls to copy_{from,to}_user would have to be
if (copy_to_user(a, b, s) != 0)
or > 0 to follow the style of the above mentioned ifs to 100%. But
that's nitpicking. :-)
--
Stefan Richter
-=====-=-==- =-=- -=---
http://arcgraph.de/sr/