Re: [mingo@elte.hu: [git pull] headers_check fixes]

[Vegard Nossum <vegard.nossum@gmail.com>]

On Wed, Jan 28, 2009 at 12:31 AM, Ingo Molnar <mingo@elte.hu> wrote:
> It would be much better if we exported _much_ less and reduced our
> cross-section to user-space. Also, the include/linux/Kbuild rules are all
> but transparent: it would also be nice if whatever we exported was be
> visible straight in the header itself, to make it obvious to people who
> modify/extend those files that those definitions are going to be exported
> to user-space.
>
> Some __user_export tag on structures perhaps? I have no good ideas here -
> #ifdefs are ugly and tags obscure the purity of the code.

Something that might or might not be doable in practice, but at least
it's a suggestion that I haven't seen elsewhere:

Create an include/user/ directory that contains a "mirror" of the
include/ directory _structure_, so that random exported header
include/linux/foo.h now has two parts -- include/linux/foo.h and
include/user/linux/foo.h.

- include/user/linux/foo.h contains the definitions that are needed by
both kernel and userspace
- include/linux/foo.h contains the definitions that are needed only by
the kernel
- include/linux/foo.h can simply #include <user/linux/foo.h> and no
other change (to source files which _use_ this header) is necessary
- no dependency on a kernel header will exist in a "user" header --
that's how it is now, but this way is more explicit
- the whole include/user/ can be shipped verbatim to /usr/include (or
wherever it is needed)
- no #ifdef __KERNEL__ or #ifdef CONFIG_ stuff in the "user" headers;
no stripping or unifdefing is needed
- it's easier to see exactly what is being exported

Of course, obvious disadvantages are:

- less readable in the sense that what used to be in one file is now
spread across two
- the split itself would probably require a tremendous effort
- other things?

It's just an idea...


Vegard

-- 
"The animistic metaphor of the bug that maliciously sneaked in while
the programmer was not looking is intellectually dishonest as it
disguises that the error is the programmer's own creation."
	-- E. W. Dijkstra, EWD1036