xorg-server 1.1.0 has AIGLX built by default: --enable-aiglx Build accelerated indirect GLX (default: enabled) For some people experimenting with Xgl, and others who have newer Intel graphics chipsets not fully supported by the open source Intel driver, this has caused problems as reported in the forums. A feature like this should really be controlled by a USE flag. Given how new this feature is, it should be a conscious choice to use AIGLX rather than a hidden source of potential problems.
What problems, exactly? Pointing to specific forums entries would be sufficient.
(In reply to comment #1) > What problems, exactly? Pointing to specific forums entries would be > sufficient. Sure: http://forums.gentoo.org/viewtopic-p-3386478-highlight-aiglx+xgl.html#3386478 http://forums.gentoo.org/viewtopic-p-3364077-highlight-aiglx+xgl.html#3364077 http://forums.gentoo.org/viewtopic-t-468194-highlight-aiglx+xgl.html http://forums.gentoo.org/viewtopic-p-3334681-highlight-aiglx+xgl.html#3334681 I seem to recall a couple of others buried in the multi-part XGL thread, but those are the ones I could find fairly quickly. Aside from any problems, this really seems like the kind of feature that USE flags were made for.
I disagree that this is the sort of feature USE flags are for. It can already be disabled at runtime in xorg.conf in the ServerFlags section (Option "AIGLX" "off"), and it's only relevant on indirect rendering, which would otherwise use software.
(In reply to comment #3) > It can already > be disabled at runtime in xorg.conf in the ServerFlags section (Option "AIGLX" > "off"), and it's only relevant on indirect rendering, which would otherwise use > software. > The DRI module can be enabled/disabled at runtime with a "Load" statement. Yet there is a dri USE flag (which I think is only used by xorg-server) to control building it. If someone is running one of the lighter WMs on older hardware and doesn't want the eye-candy, shouldn't they have the option of not building AIGLX? For that matter, when Xgl is officially supported in xorg-server why should an Xgl user want unneeded AIGLX code built, or vice versa? There may also be a need to have this USE flag for other packages as they become "aware" of the different servers/compositing WMs. By the way, I didn't know about that ServerFlag, so thanks for that info. That makes solving problems easier. ;-)
(In reply to comment #4) > The DRI module can be enabled/disabled at runtime with a "Load" statement. Yet > there is a dri USE flag (which I think is only used by xorg-server) to control > building it. If someone is running one of the lighter WMs on older hardware and > doesn't want the eye-candy, shouldn't they have the option of not building > AIGLX? For that matter, when Xgl is officially supported in xorg-server why > should an Xgl user want unneeded AIGLX code built, or vice versa? There may > also be a need to have this USE flag for other packages as they become "aware" > of the different servers/compositing WMs. AIGLX isn't purely about eye candy. It makes running accelerated applications across the network possible (they use indirect rendering, which AIGLX accelerates). It makes eye candy possible, but it doesn't provide it itself.
I've been running Xgl for a couple of months now so I know that there are also major usability improvements made possible by this technology, not just eye-candy. My point still stands, though -- there are many useful features that I would not want to give up, but that others choose to not build in order to have a leaner system, or a more secure system, or fewer problems (real or potential). Given that adding a USE flag for AIGLX is simply exposing the existing configure option, why not give those users this flexibility?
Because there aren't good enough reasons to disable it. It doesn't pull in any extra dependencies and the size difference to libglx is negligible. Choice for the sake of choice is not a reason. We (Gentoo) want to give you choice when choice is useful, not overwhelm you with largely irrelevant, useless or ineffectual choices.