Dear Shawn Guo,

On Tue, May 01, 2012 at 04:02:56PM +0200, Marek Vasut wrote:

quoted

Dear Shawn Guo,

quoted

On Mon, Apr 30, 2012 at 11:01:08PM +0200, Wolfram Sang wrote:

quoted

quoted

quoted

quoted

It complains that data might be undefined, do you want it in a
separate patch then ?

Yes, but only if you can prove that the compiler is right.

It's not right, because that variable is always initialized, but
this at least squashes the compile warning.

The compiler needs to be fixed, not the kernel.

I second this.

I agree on this too.

quoted

quoted

Heh, this is coming up again.  I'm not convincing you to take the
change, but just curious what if this is not a compile warning but
an error for the same cause that compiler is not right.

Can you elaborate?

Will we have the kernel not compilable for a few cycles until the
compiler fixing is available, or we work it around to keep kernel
compiling and working before we have the compiler fixed?

This is kind of lacmus paper showing if the compiler was fixed. On the other 
hand, many people won't have a fixed compiler for a long time, so I'd be for 
applying this patch.

Best regards,
Marek Vasut