Re: Sanitise ethtool.h and mii.h for userspace.

[David Woodhouse <dwmw2@infradead.org>]

On Tue, 2006-06-20 at 21:18 -0400, Jeff Garzik wrote:
> How can reviewers make an informed decision, when you completely failed 
> to note:
> 
> * This breaks the primary userspace user of this header, ethtool(8)

I cannot reproduce with either an ethtool-3 tarball or a fresh checkout
from your git tree. Can you show me the error?

It's a somewhat surprising allegation, because ethtool doesn't even
_use_ the header directly from the kernel. It has its own copy, and
currently includes it like this...

typedef unsigned long long u64;         /* hack, so we may include kernel's ethtool.h */
typedef __uint32_t u32;         /* ditto */
typedef __uint16_t u16;         /* ditto */
typedef __uint8_t u8;           /* ditto */
#include "ethtool-copy.h"

> * The patch was NAK'd (and I don't even get a "Naked-by:" header :))

Sorry, my recollection was that you backed down after everyone turned on
you and declared that "but I _want_ to use 'u32' for userspace stuff
because underscores hurt my eyes" was too silly for words. You didn't
get a mention in the commit comment because I'd already committed it by
the time we had the discussion. But I did mention it to Andrew at the
time I asked him to put the hdrcleanup tree into -mm, and he seemed
perfectly happy with the change.

> * Despite knowing all this for quite some time, no associated userspace 
> fix patch has ever appeared.

That's because to the best of my knowledge, userspace doesn't _need_ any
fix at the moment. We've built the whole of the Fedora Core 6 test 1
release against (a subset of) these headers, and that includes ethtool.

> If you are going to break stuff, AT LEAST TELL PEOPLE IN ALL CAPS ABOUT 
> IT, rather than providing the highly deceptive description as above. 
> And be courteous enough to help fix the breakage, if you please. 

If I break stuff, I promise I'll bear that in mind.

-- 
dwmw2