This is the mail archive of the crossgcc@sourceware.org mailing list for the crossgcc project.

See the CrossGCC FAQ for lots more information.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: a couple thoughts on ct-ng


On Wed, 11 Apr 2007, Yann E. MORIN wrote:

> On Wednesday 11 April 2007 15:41, Robert P. J. Day wrote:

> > 1) there's no dependency for the choice of endianness on the
> > selected architecture.  does that really make sense?  shouldn't
> > some architectures *enforce* the selection of endianness?
>
> Effectively, there's none. I've got some Kconfig fu to test tonight
> to try big/little endian selection dependent on the architecture.
> Work for tonight...

basically, that would be implemented by the "select" Kconfig
directive, no?  pretty straightforward.

> > 2) the location of the kernel headers should be extended to
> > include a simple directory, in which i would prefer to place my
> > *own* sanitized kernel headers generated by running a "make
> > headers_install" in my kernel source tree of choice.
>
> I see. That could be an option to provide "custom directory
> headers", alongside with the other options: "copied", "sanitised" or
> "installed". Comments?
>
> Now, why would you want your own kernel headers? Aren't those from
> the pristine kernel enough? I can't see a practical situation where
> the headers from a modified kernel would change from those of a
> pristine kernel... But I might miss some corner cases... :-)

i'm not sure i know what you're asking here.  this would just be a way
to verify that the sanitized headers produced by the latest kernel
source tree are still compatible.  and, trust me, those headers
produced by running "make headers_install" *do* change on a regular
basis.  i should know -- i've submitted a few cleanup patches for
them.

> > 3) regarding the choice of "kernel to use", what is the value of this
> > beyond the need to get the sanitized headers?  if i provide my own
> > headers, is there any reason to need a kernel source tree beyond that?
> > if not, then there should be a choice of "none."  in fact, if all i
> > need are those headers, i would really have no need of that entire
> > "kernel" config submenu, would i?
>
> There are two questions there:
>  - "kernel to use" is the OS we target in the compiler: Linux, Cygwin (or
>    others to be added later).
>  - the kernel headers are those headers needed to build against the OS selected
>    above. So if you have your own kernel headers (be it Linux or cygwin), then
>    the answer is: see answer to question 2, above.
>
> So definitely, you don't want to say "none" to "kernel to use". I agree it's
> misnamed, and should be "target kernel". Doing that, commit pending shortly.

or, better yet, "target OS".

> And you'd still need the kernel sub-menu to say "target kernel" =
> "Linux" and "kernel headers" = "custom directory" and "custom
> directory" = "/some/place"
>
> What do you think of that? Is my interpretation correct of your
> needs?

give me a bit, i'll go over this.

rday
-- 
========================================================================
Robert P. J. Day
Linux Consulting, Training and Annoying Kernel Pedantry
Waterloo, Ontario, CANADA

http://fsdev.net/wiki/index.php?title=Main_Page
========================================================================

--
For unsubscribe information see http://sourceware.org/lists.html#faq


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]