Re: [patch 119/197] scripts/coccinelle/free: Delete NULL test before freeing functions

[SF Markus Elfring]

> +@r depends on context || report || org @
> +expression E;
> +position p;
> +@@
> +
> +* if (E)
> +* \(kfree@p\|debugfs_remove@p\|debugfs_remove_recursive@p\|usb_free_urb\)(E);

How do you think about to add the construct "@p" also behind the function name
"usb_free_urb"?


Is my previous update suggestion for such an issue worth for another look, too?

Regards,
Markus
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majo...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

[Julia Lawall]

On Sat, 9 Aug 2014, SF Markus Elfring wrote:

> > +@r depends on context || report || org @
> > +expression E;
> > +position p;
> > +@@
> > +
> > +* if (E)
> > +* \(kfree@p\|debugfs_remove@p\|debugfs_remove_recursive@p\|usb_free_urb\)(E);
>
> How do you think about to add the construct "@p" also behind the function name
> "usb_free_urb"?

Yes, that would be needed, thanks. Also, it was recently suggested to add
of_node_put to the list of things that are checked for.

> Is my previous update suggestion for such an issue worth for another look, too?

I still don't think this should be done for any random function that
performs a null test on its argument. The corrections involved here are
not as trivial as they would seem. Often it is not the case that the null
test at the call site should be just deleted, instead the code should be
reorganized. Making a rule that treats 5000 functions that are
automatically selected will just encourage people to do sloppy things.

(Personally, I don't like the whole null test removal idea. It mixes the
notion of something that is necessary and has failed and something that is
optional. In the case of something that is necessary and has failed, the
null value is usually statically apparent, and the code can be reorganized
so that a null value doesn't go where it is not wanted. In the case of
something that is simply optional, the presence of the null test gives
the reader of the code some information. Removing this information in
rarely executed code seems unfortunate.)

julia

>
> Regards,
> Markus
> --
> To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in

[SF Markus Elfring]

> I still don't think this should be done for any random function that
> performs a null test on its argument. The corrections involved here are
> not as trivial as they would seem.

I would prefer to make the list of corresponding function names more complete.

> Often it is not the case that the null test at the call site should be
> just deleted, instead the code should be reorganized.

Which source code places do you know where a different approach might look better?


> (Personally, I don't like the whole null test removal idea. [...]

Would you like to clarify involved software concerns a bit more?

Regards,
Markus
--

[SF Markus Elfring]

>>> +@r depends on context || report || org @
>>> +expression E;
>>> +position p;
>>> +@@
>>> +
>>> +* if (E)
>>> +* \(kfree@p\|debugfs_remove@p\|debugfs_remove_recursive@p\|usb_free_urb\)(E);
>>
>> How do you think about to add the construct "@p" also behind the function name
>> "usb_free_urb"?
>
> Yes, that would be needed, thanks. Also, it was recently suggested to add
> of_node_put to the list of things that are checked for.

Would you like to achieve further improvements for this approach?
https://git.kernel.org/cgit/linux/kernel/git/stable/linux-stable.git/tree/scripts/coccinelle/free/ifnullfree.cocci?id=refs/tags/v4.0.1#n34

Regards,
Markus

--