Automatic Vesa graphics fallback

[jonas.kirilla]

Have a graphics driver realize (before crashing) that it has failed setting up the hardware, have it abort and fall back to vesa, letting the vesa driver have a shot.

Is it possible and worthwhile?

[anevilyak]

Depends on what you mean by "before crashing". Generally the drivers already do error check initialization and don't publish a node in /dev/graphics if they fail, (as a result of which the vesa driver will obviously be selected). If you mean the scenario where they crash the app_server, then I don't realistically see how, since most of the time those forms of crashes are just a case of the same pointer/memory errors that result in every day app segfaults, which aren't realistically detectable in all cases, and should simply be fixed in the driver in question.

The other scenario is where things lock up while doing the initialization, which is also often not really detectable/avoidable if it's a consequence of the graphics hardware itself corrupting something rather than a fault in the driver, except by determining that's the issue and blacklisting or doing a quirks workaround for that piece of hardware after the fact, as is already done in various cases (not just graphics).

[kallisti5]

Actually, if the accelerant fails to load (even returning cleanly), the app_server crashes. (the radeon_hd accelerant did this if I returned on an error)

[diver]

Hasn't that been fixed recently?

[waddlesplash]

Yes. Problems with fallback are now with individual drivers and not app_server. At least radeon_hd falls back correctly in some cases.