Re: Gnumach: spl.h vs machspl.h

[Thomas Bushnell BSG]

> i386/i386/spl.h
>     #include ipl.h
> 
> i386/i386/ipl.h
>     #ifdef __ASSEMBLER__
>     #include <i386/machspl.h>
> ...

You had me all confused looking at your email.  What I see in
<i386/ipl.h> is this:

#ifndef ASSEMBLER
#include <machine/machspl.h>

> First question is, is this an issue?  machspl.h doesn't seem to have any 
> #ifdef _I386_MACHSPL_H_ wrapper around it.

It hardly needs one, since it's just another name for <.../spl.h>

What happened here seems to be the following:

Once there was <i386/ipl.h>, which defined the values of the different
priority levels.  There was also <i386/machspl.h>, which defined some
other stuff.

Then OSF came along, and wrote <i386/spl.h>, which replaced the contents
of <i386/machspl.h>, so the latter was truncated to just include the
former.  Meanwhile, because (drivers?) were accustomed to include spl.h,
it was changed to include ipl.h back.  Ain't that fun?

It seems to me these can all be one file.  Remember to protect the
assembly-improper definitions with #if(n)def ASSEMBLER.

However, if there are drivers including spl.h, this may have been the
reason for OSF's change.  I would suggest examining the changelogs and
reconstructing the history, and only then making a change.

Thomas

signature.asc
Description: This is a digitally signed message part