Re: [gdm-list] GDM D-Bus on Solaris

[Brian Cameron <Brian Cameron Sun COM>]

Jon:

Sorry for the follow-up, but Artem mentioned to me the following:

---

Do 'prtconf -v' and look for the "display" node. You will see properties like these:

            display, instance #0
                    name='model' type=string items=1
                        value='VGA compatible controller'
                    name='subsystem-vendor-id' type=int items=1
                        value=000010de
                    name='subsystem-id' type=int items=1
                        value=00000168
                    name='device_type' type=string items=1
                        value='display'
                    name='unit-address' type=string items=1
                    name='class-code' type=int items=1
                        value=00030000
                    name='revision-id' type=int items=1
                        value=000000a2
                    name='vendor-id' type=int items=1
                        value=000010de
                    name='device-id' type=int items=1
                        value=0000018a

As you can see, we already map some of these to HAL's pci.* properties. The 
'class-code' above corresponds to pci.device_class. So the patch would be 
something like:

>       if (di_prop_lookup_ints (DDI_DEV_T_ANY, node, "class-code", &i) > 0) {
>              class = i[0];
>          }
>          hal_device_property_set_int (d, "pci.device_class", class);

If you can confirm that that's what your code needs, I can provide you with a 
test HAL binary. It's an easy fix. If what you need is more elaborate, then most 
likely I won't have time for that.

---

So, Jon, do you think this would fix the problem?  If so, I can get a test
binary from Artem and see if we can get this fixed.  Based on this, are there
any other properties that we need to be concerned about?  Looking at the code,
I only see the pci.device_class property being referenced.

Brian


> > The error clobbering is now fixed.
>
> Thanks.
>
> > So, what the code is trying to do is detect local "seats" by looking
> > for graphics cards.  Also see:
> > http://lists.freedesktop.org/archives/hal/2007-July/009160.html
> >
> > Should be using a different approach with Solaris HAL?
>
> According to Artem, who has worked on HAL in Solaris, he says that he
> doesn't see graphics cards in the HAL spec, and PCI graphics cards will
> appear in the list as a generic PCI device.
>
> It seems that the pci.device_class property is missing on the Sun
> implementation of HAL.  He says that the fix might be as simple as
> adding this property.
>
> Or is it possible to get the needed information another way on
> Solaris?  Perhaps HAL could check if the property exists, and if not,
> have fallback code for Solaris?
>
> Brian