[Intel-gfx] [PATCH V1] drm: export get and put connector type id
ling.ma at intel.com
Thu May 28 09:11:33 CEST 2009
>From: Eric Anholt [mailto:eric at anholt.net]
>Sent: Wednesday, May 27, 2009 12:35 AM
>To: Ma, Ling
>Cc: dri-devel at lists.sourceforge.net; intel-gfx at lists.freedesktop.org
>Subject: RE: [Intel-gfx] [PATCH V1] drm: export get and put
>connector type id
>On Tue, 2009-05-26 at 11:12 +0800, Ma, Ling wrote:
>> >-----Original Message-----
>> >From: Eric Anholt [mailto:eric at anholt.net]
>> >Sent: Saturday, May 23, 2009 3:12 AM
>> >To: Ma, Ling
>> >Cc: dri-devel at lists.sourceforge.net; intel-gfx at lists.freedesktop.org
>> >Subject: Re: [Intel-gfx] [PATCH V1] drm: export get and put
>> >connector type id
>> >On Fri, 2009-05-22 at 20:15 +0800, Ma Ling wrote:
>> >> We use connector type id as sequence to avoid duplicate
>> >connector name.
>> >> However for some connectors that advertise multiple functions
>> >> we have to meet the same issue, i.e. connector advertise two
>> >> DVI and VGA, actually only VGA is real one, in init function
>> >> is allocated for CRT, and card0-DVI-D-1 is allocated for
>> >DVI, but until
>> >> detection process complete we could be aware of VGA, not DVI
>> >for this connector,
>> >> then we change constructed connector type as VGA, which cause that
>> >> the name card0-VGA-1 would be allocated to both a CRT and
>> >that connector.
>> >> The patch intends to avoid above case by exporting get and
>> >put connector type id.
>> >This whole renaming connectors thing seems broken to me. Whatever
>> >detection is required should be done once at init, and appropriate
>> >connectors created. Is there something preventing that?
>> We determin real output by response from sdvocommand.
>> At init if we don't plug any output, the response from sdvo
>device is 0,
>> so we have to chose real output untile detection complete.
>In this case, I think we should expose the maximum reliable information
>at init time (even if it means "unknown"), and not change it. Changing
>the connector names at runtime seems bad to me. It'll only confuse
>people and their xrandr scripts.
It is good idea :), I will expose all possible connectors at init time, delete redundant ones and leave the real connector at detection time.
>eric at anholt.net eric.anholt at intel.com
More information about the Intel-gfx