Dmitry Torokhov wrote at Tuesday, January 31, 2012 12:48 PM:

On Tue, Jan 31, 2012 at 11:29:35AM -0800, Stephen Warren wrote:

quoted

Dmitry Torokhov wrote at Tuesday, January 31, 2012 12:09 PM:

quoted

On Thu, Dec 29, 2011 at 07:25:59PM -0800, Dmitry Torokhov wrote:

quoted

On Fri, Dec 09, 2011 at 03:17:54PM -0800, Stephen Warren wrote:

...

quoted

quoted

quoted

quoted

However, I suppose if we were to change the structure of the pdata to be:

struct tegra_kbc_pin_cfg {
    bool is_row;
    u8 row_col_id;
    u8 pin_id;
};

Then struct tegra_kbc_platform_data could indeed have a pointer to a
variable-sized array of these, which would avoid the "en" member.

How about we change bool to enum instead, like in the patch below?

...

quoted

I'd rather PIN_CFG_IGNORE was the first enum, so it gets value 0, or this
probably won't solve the original issue?

The benefit of current definition is that it is compatible with old
ones using bool...

There are no in-tree users of this type yet, so compatibility isn't
really an issue. And if we don't make that change, anyone who does start
to use it is going to have to initialize every element of the array even
when they aren't used.

-- 
nvpublic