Saturday, April 09, 2016
Well, finally found some time and the source of my old website back. It's online again, the links here are updated to point there (www.rudolfs-place.nl).
I'll update that old site a bit so all (or most) links will work correctly again. I plan to start posting a bit every now and then again, though time is limited due to busy real life.
In the meantime I've done a few smaller things:
- Updated the nvidia driver for basic dualhead support with the native Haiku screenprefs app;
- Added a new setting to nv.settings to disable use of EDID (item check_edid) so bad-behaving monitors can be used on higher resolutions as well;
- Worked on the VIA driver. It now has screen and overlay support for K8M800. Also it supports more mainboard memory types now (DDR166 and DDR200). It's back in the Haiku image as well now;
- Fixed overlay generally in app_server for 8, 15 and 16 bit color depths.
Other work done:
While working again on the drivers I encountered, in my head, strange behaviour from the Haiku Backgrounds app. The color of the color selection sliders changes while moving the sliders around. This was not the case on BeOS. Since I was focussed on driver issues and I always use this app from that perspective I concluded that there might be something wrong with the driver(s) causing this behaviour.
In the end it turned out that it's actually new 'intended behaviour' so I can leave it alone. But not before I encountered a small error in the underlying class BColorControl with doublebuffering. This (along with other things) was fixed in the meantime by John Scipione so that class is now fully working OK from driver testing perspective.
Next up I am busy with trying to locate the exact problem Haiku's app_server currently has with hardcursors: while the drivers I wrote support this function, it still doesn't work on Haiku. It turns out Haiku has a new driverhook for a new bitmap style cursor that more modern graphics card support (like nVidia).
I need to run a manually compiled and installed app_server on my system, but I can't get this going (even though Axel gave me a few helpfull pointers). Maybe Haiku behaves different than it was meant to be? Don't really know yet.
I do know that indeed app_server has a fault for the cursors as I _can_ get it going if I fake a new style cursor by exporting and accepting the new hook. If that hook isn't there or fails app_server incorrectly assumes the cursor does not work and falls back to software cursor.
Anyhow.. still fiddling with this one..
In the meantime I've been reading up on the VBE 3 (VESA BIOS extensions) specs to see if it would be possible to add widescreen support to our VESA graphics fallback driver. Officially this can't be done I fear. This I did not expect, I'm disappointed in that. I had hoped you could set a custom 'modeline' but that cannot be done: only custom timing (refresh rates) for existing resolutions can be set.
Anyhow I'll do some tests to see if it's possible after all: there's a function concerning panels native resolutions: that would be the way to go with VESA, if possible.
No tests done yet though.
So, first I am still busy with the cursor. Also I am going to do some tests with virtual screens which depends on availability of a hardcursor (at least on BeOS). Since Haiku's screenprefs panel makes 'illegal' use of that for dualhead support on ATI cards I fear this won't work correctly at all times yet either..
That's it for now.