[PATCH xserver] Fix Xorg -configure not working anymore

[Hans de Goede]

Hi,

On 08/16/2016 06:36 PM, Adam Jackson wrote:
> On Mon, 2016-08-08 at 14:53 +0200, Hans de Goede wrote:
>
>> Adding support for the xf86Configure code to xf86platformBus.c
>> is non trivial and since we advise users to normally run without
>> any Xorg.conf at all not worth the trouble.
>
> I'm guessing "non-trivial" is used here in the "I don't like this code
> so I don't want to touch it" sense.

I don't want to touch it is part of it, I did actually look at what
needed to be one to extend the platform bus code to work with
the xf86Configure code. There are 2 problems:

1) The xf86Configure code seems to have 2 stages, one enumeration
stage and one to actually bind drivers, implementing the enumeration
stage seems non trivial

2) The xf86Configure code also expects a callback to check if a device
has been enumerated already to avoid adding devices twice. The PCI
code uses an internal list for this, if we add support for this to
the platform code, it would need to check both a list of devices
it has itself, as well as hook into the pci-devices list for pci
devs. Again non trivial

> Which, fair enough I guess, but I
> really don't think it's that difficult to add. Like, pretty sure one
> conditional at the bottom of doPlatformProbe would be about all you
> need.

I don't think so, see above.

Either way we need to fix this, my change at least stops the regression
where this no longer works at all for any drivers which have added
platform bus support (pretty much all of them). So unless someone
volunteers to fix this properly, then I suggest we go with my fix.

> Also: "we advise users not to use xorg.conf" is a bad meme, can we
> please stop. It is still the config file. If you need it, you need it.

Right, note I wrote "advise users to *normally* run without any Xorg.conf"

>
>> However some users still want to use Xorg -configure to generate a
>> template config file, this commit implements a minimal fix to make
>> things work again for PCI devices by skipping the platform
>> probe method when xf86DoConfigure is set.
>
> This means you'll fail to generate a config file under Hyper-V, where
> the fbdev device is platform not pci.

Something which has _never_ worked, as said currently we have a regression
with Xorg -configure no longer working anywhere at all. So unless someone
steps up to fix this properly I suggest we go with my patch to at least
revert back to the old behavior and not regress.

Regards,

Hans