* Martin Husemann <martin%duskware.de@localhost> wrote: > On Fri, Feb 03, 2006 at 05:10:19PM +0100, Ed Schouten wrote: > > the CPU quirk workarounds and the vector instructions), but the userland > > is built with a regular mipsel toolchain > > Yes, you are right. It would be *realy* cool to get the EE patches lifted > into (a) official gcc anytime, and (b) into a 4.1-branch gcc soon, so we could > use them as local patches in the NetBSD tree (once a newer gcc is imported). So how about sending in a small patch to the GNU folks that fixes the userland crashes for starters? I guess compiling the NetBSD kernel with a separate toolchain isn't that bad. It sure needs to be resolved finally, but getting a working userland does have a higher priority ;-) In the other news: * Ed Schouten <ed%fxq.nl@localhost> wrote: > I also found some other small bugs: > > - The gsfb driver shows some small corruption on the first eight rows > and two columns (16 chars) of the display; a lot of the time the > characters are painted black. Most of the time when scrolling upwards. After some debugging: there are actually two bugs in the gsfb driver: - In the function `gsfb_set_cursor_pos', there's an off-by-one bug which draws the cursor at 7x15 instead of 8x16 (can be fixed by removing the '- 1's from the source. Anyone with a commit bit in here? ;-) - The `gsfb_cursor_cmd' array contains some faulty data which garbles up the top rows when repainting the cursor. I don't have a clue how the Playstation 2 video chip works, but I know the bug is there - when I disable the entire cursor code, my terminal works 100% correctly. Yours, -- Ed Schouten <ed%fxq.nl@localhost> WWW: http://g-rave.nl/
Attachment:
pgpKLeqlI94DL.pgp
Description: PGP signature