committing the wsfb driver

Matthieu Herrb matthieu.herrb at laas.fr
Sat Sep 25 10:52:49 PDT 2004


Adam Jackson wrote:
> On Saturday 25 September 2004 06:17, Matthieu Herrb wrote:
> 
>>Are there any objections?
> 
> 
> This looks bad:
> 
>         if (!osname || strcmp(osname, "openbsd") != 0) {
>                 if (errmaj)
>                         *errmaj = LDR_BADOS;
>                 if (errmin)
>                         *errmin = 0;
>                 return NULL;
>         }
> 
> How does this work on NetBSD?

By adding a test for NetBSD here... The cvsweb you're seeing is 
OpenBSD's one and since there is no central CVS repository for this code 
  yet, I've not merged NetBSD changes yet.

> 
> It's nice to be able to compile all drivers on all OSes for various reasons 
> (tinderboxing, dlloader sanity checks) even if they won't run correctly on 
> all OSes.

I don't really agree. There are already a couple of drivers built only 
on Linux. And I don't see how using fake headers or stub functions  to 
build a driver that cannot run on a system provides any benefit.

Tinderbox is there to make sure the tree builds natively on a wide range 
of machine.  Energy must be thrown at it to keep it running on many 
different systems and ensuring that the results are looked at.

   AFAICT the "#include <dev/wscons/wsconsio.h>" is the only major
> problem.  I'd like to change that to include a wsconsio.h from, say, 
> os-support/, which then imports the real one on Net or Open or fakes the 
> necessary defines for anything else.  Does that sound reasonable?

I'd rather avoid this kind of duplication. (there is a precedent with 
libusbhid, and it has been a pain to maintain).
-- 
					Matthieu



More information about the xorg mailing list