Please see bug #414853 comment 5 for why we need something like this. https://bugs.gentoo.org/show_bug.cgi?id=414853#c5 Thank you!
I think that making Xvfb style parameters the external interface of the eclass might not be a good idea. We may have to replace Xvfb with something else in the future (bug 409925).
Does that mean that xf86-video-dummy does not allow controlling the resolution? If it does, It would be no problem, right?
It allows controlling the resolution, but not in the way that it is done with Xvfb. So in its suggested format, ${VIRTUALXRES} would need to be parsed and converted to xorg.conf syntax.
How about three variables then to save any parsing by preserving the structure: VIRTUALX_RES_WIDTH or _X VIRTUALX_RES_HEIGHT or _Y VIRTUALX_RES_DEPTH or _BITS Would that work?
I increased the Xvfb resolution to address the immediate problem in bug 414853. When I add xf86-video-dummy support to virtualx.eclass, I will include a way to specify resolution, possibly the one from comment 4.
(In reply to comment #4) > How about three variables then to save any parsing by preserving the > structure: > > VIRTUALX_RES_WIDTH or _X > VIRTUALX_RES_HEIGHT or _Y > VIRTUALX_RES_DEPTH or _BITS > > Would that work? Personally my mind screams KISS when I see that. As most (all?) RANDR1.2 drivers creates a set of modes with names in the format <width>x<height> I think we only need on variable (VIRTUALX_RES?) that takes one parameter of that format. This is the same as Option "PreferredMode" in xorg.conf usually wants as a parameter, and the same goes for xrandr. With other word, what I propose is something in the style of VIRTUALX_RES="1280x1024" which could be parsed directly into a Option "PreferredMode" "1280x1024" in xorg.conf or set with xrandr --output <something> --mode "1280x1024" At the same time I start to think that virtualx always should unset DISPLAY. I am somewhat tired of getting my session highjacked by a testcase when I do a "emerge -e world" in the background....
xf86-video-dummy does not support xrandr-1.2 yet.
We haven't found another need for this parameter in more than six years.