GNU Mach source code (was: SiS900)

[Thomas Schwinge]

[ CCed to bug-hurd.  Please reply where it is appropriate. ]

On Sun, May 01, 2005 at 04:08:05PM +0200, Alfred M. Szmidt wrote:
>    You should put Linux's original files (i.e. the versions your patch
>    is based on) into linux/src/ and put your modified files into
>    linux/dev/.  The files from linux/dev/ will shadow the linux/src/
>    ones.
> 
> I disagree very strongly (and thus disagree with the comment in one of
> the Makefiles if it is to be followed blindly).

I agree to your statements.
I was merely quoting and trying to interpret that Makefile and the
"common practice" I saw so far (not that much, though ;-).

> Code that is hacked
> to work specifically for Mach and/or the Hurd should be put into
> linux/dev, but code that has been only modifed to add a PCI ID,
> updated or has a bug fixed and contains no Mach/Hurd specific code
> should always go into linux/src.

However, later it must still be obvious to figure out easily if a
committed change is an update from Linux-2.0.x, a backport from e.g.
Linux-2.2.x, or a Mach-specific change.  The ChangeLog is not always
explicit there, at least IMHO.

> If we follow the "rule" that you note we will have lots of moving
> files back and forth for no apparant reason, and thus making things
> impossible to follow.  Imaging the following scenario [...]

That would be less of a problem, if the revision control system
supported file renames, etc., which the currently used one doesn't.


On a related side-note:
I'm currently bootstrapping the Glorious Glasgow Haskell Compiler
<URL:http://www.haskell.org/ghc/> on GNU/Hurd to be able to build darcs
<URL:http://darcs.net/> to be able to use my revision control system of
choice.


Regards,
 Thomas