Re: [PATCH 1/3] spi: Add helper functions for setting up transfers

[Julia Lawall <julia.lawall@lip6.fr>]

On Thu, 10 Jan 2013, Lars-Peter Clausen wrote:

> On 01/10/2013 09:53 AM, Julia Lawall wrote:
> >> +@r1@
> >> +identifier fn;
> >> +identifier xfers;
> >> +@@
> >> +fn(...)
> >> +{
> >> +	...
> >> +(
> >> +	struct spi_transfer xfers[...];
> >> +|
> >> +	struct spi_transfer xfers[];
> >> +)
> >> +	...
> >> +}
> > 
> > Can it happen that there would be more than one spi_transfer or spi_message
> > variable per function?  This semantic patch will only treat the case where
> > there is only one, because the ... before an after the variable declaration
> > won't match another declaration of the same form.
> > 
> > julia
> 
> I guess it could happen, but I would consider it to be very rare. There are
> a few examples of multiple transfers in the kernel. But most of them look like
> 
> struct spi_message msg;
> struct spi_transfer xfer_foo;
> struct spi_transfer xfer_bar;
> 
> ...
> spi_message_add_tail(&xfer_foo, &msg);
> spi_message_add_tail(&xfer_bar, &msg);
> 
> So the transformation can't be applied here anyway.
> 
> Do you have an idea how to change the rule to work with multiple
> transfers/messages per function? If it would make the cocci file more
> complex I wouldn't bother to take care of it, since it basically has no
> practical use.

Probably the simplest thing is to put when any on all of the ...s
It might get slower, though.

Alternatively you could have a rule at the end that prints a warning for 
any cases that are not transformed.

julia