[PATCH] Improved autoconfig drivers matching

[Peter Hutterer]

On Mon, May 01, 2017 at 06:03:22PM +0900, Michel Dänzer wrote:
> On 28/04/17 04:11 AM, Adam Jackson wrote:
> > On Wed, 2017-04-12 at 23:42 +0100, Emil Velikov wrote:
> >>> On 12 April 2017 at 23:05, Aaron Plattner <aplattner at nvidia.com> wrote:
> >>> On 07/12/2016 04:31 PM, Emil Velikov wrote:
> >>>> Since xf86platformBus.h is part of the SDK, If we do this, then the
> >>>> new header must become one as well (should be listed in sdk_HEADERS).
> >>>> Alternatively we can forward declare XF86MatchedDrivers and include
> >>>> the header in EXTRA_DIST. Not sure if the latter is a good idea
> >>>> though, since the actual ABI will be undefined/private.
> >>>>
> >>>> Or better yet, neither of the two exported symbols
> >>>> (xf86PlatformDeviceCheckBusID, xf86PlatformMatchDriver) is used and
> >>>> imho we can remove them. Seems that the header is used solely for the
> >>>> ODEV management, which isn't platform devices specific and one can
> >>>> just move those parts into a separate header and use _it_ in the SDK ?
> >>>>
> >>>> But all that (everything but the sdk_HEADERS/EXTRA_DIST fix) is added
> >>>> bogus, which shouldn't stop the patch from landing.
> >>> Another customer ran into this recently. Adam, can this be merged? I don't
> >>> think Emil's reply was a nack.
> >>
> >> Precisely. My earlier message should have read:
> >>
> >> xf86MatchDrivers.h must be in the sdk_HEADERS or you'll need build
> >> hacks in each driver. With that the patch is
> >> Reviewed-by: Emil Velikov <emil.l.velikov at gmail.com>
> > 
> > Fixed up (and rebased and made meson-aware) and merged:
> > 
> > To ssh://git.freedesktop.org/git/xorg/xserver
> >    1549e3037..112d0d7d0  master -> master
> 
> This broke things on my development box.

broke things for me too, reverting fixed it.

May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (==) Matched intel as autoconfigured driver 0
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (==) Matched modesetting as autoconfigured driver 1
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (==) Matched fbdev as autoconfigured driver 2
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (==) Matched vesa as autoconfigured driver 3
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (==) Assigned the driver to the xf86ConfigLayout
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (II) LoadModule: "intel"
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (WW) Warning, couldn't open module intel
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (EE) Failed to load module "intel" (module does not exist, 0)
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (EE) No drivers available.
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (EE)
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: Fatal server error:
May 04 11:38:22 jelly /usr/libexec/gdm-x-session[28936]: (EE) no screens found(EE)

i.e. it now fails hard instead of continuing with the next driver.
 
Cheers,
   Peter