Please test: use /dev/mem instead of /dev/console for memory-mapping framebuffers in X11 on PowerPC

Wed Jul 23 04:43:10 UTC 2014

I tried two general types of contexts: G5/PowerPC64 and G4/PowerPC, all PowerMacs. Unfortunately I could only complete the G5 tests (no problems found). The G4 context had problems without the patch and so did not provide a known-good context to compare to --and I ran out of time for getting past that.

Xorg + xfce4 basic context, xorg.conf as produced by Xorg -configure for the video hardware except as explicitly noted below, startxfce4 used to see what would happen.

G5's...

The SSD was moved between the G5's and so only one build was involved, but separate Xorg -configure's were used.

uname -a : FreeBSD FBSDG5S0 10.0-STABLE FreeBSD 10.0-STABLE #0 r268571: Sun Jul 13 06:28:19 UTC 2014     root at grind.freebsd.org:/usr/obj/powerpc.powerpc64/usr/src/sys/GENERIC64  powerpc.

Basic use for the PowerMac G5's (PowerPC64) did not show any problems with applying the patch and rebuilding what was dependent. It ended up working the same as before the patch from what I could see. Two variants used:

PowerMac G5 (7,11), GeForce 7800 GT

PowerMac G5 (7,2), Radeon 9800PRO NH (AGP)

(That last already had NoAccel enabled so that text displays correctly --and still does for the patch test: bug-for-bug compatible, bug-for-bug matching workaround.)

Other than needing to use NoAccel for the Radeon: the original and updated builds for the above went fine. The usage seemed to match before vs. after the patch.

G4's...

Unfortunately I'm not going to be able to be a test case for the patch for the G4's (PowerPC). Details follow for why I could get that far in a straight forward/quick manor.

Example uname -a : FreeBSD FBSDG4S0 10.0-STABLE FreeBSD 10.0-STABLE #0 r268571: Sun Jul 13 05:15:31 UTC 2014     root at grind.freebsd.org:/usr/obj/powerpc.powerpc/usr/src/sys/GENERIC  powerpc

Two separate SSDs and so 2 separate builds involved for the G4's. Swapping SSD's and swapping the xorg.conf's to match the video HW did not change the behavior below. The G4's and their video are working fine for Mac OS X 10.5.

I had not built Xorg or xfce4 for the G4's yet and it takes a while, using most of the time I had left for this activity before I'm away from the PowerMacs for a significant time (again). Once built the original (un-patched) builds did not work for either G4 context...

PowerMac G4 (3,6), GeForce4 Ti 4600: UI hangs during the initial xfce4 "welcome" screen update, frequently without the background being finished. What is displayed seems fine as far as it got. Can still Option-Fn just fine to get back to VTn and use it.

PowerMac G4 (3,6), ATI Radeon 9000/PRO If (AGP/PCI): Immediate random, varying garbage displayed, not a static image: complete loss of a coherent or even stable image. Stays that way. Unable to Option-Fn. I tried typing blind but got no evidence that it did anything.

I do not have the time to investigate or get past these issues before I'll be away from these PowerMacs for a long time. The Radeon the log file for the above is essentially the same as it produced for Xorg -configure for the Radeon. The two times in [...]'s for the below are 0.001 apart in each example I've seen and a bunch of the prior lines had the same [...] as the first shown below in each case. No (EE)'s or the like:

[...] Number of create screens does not match number of detected devices.
  Configuration failed.
[...] Server terminated with error (2). Closing log file.

I'm afraid that I need to stop on this activity with that much information for a significant time. Sorry.

Side note: Earlier when I experimented with buildworld (nothing changed and not installed, just built) the build times were about:

Each PowerMac G5 (7,11), 2.5GHz quad, 16GB, -j 6: 2 hr (add about 0.5 hr for buildkernel)
PowerMac G5 (7,2), 2.0GHz dual, 12 GB, -j 4: 4 hr (add about 1 hr for buildkernel)
Each PowerMac G4 (3,6), 1.42GHz dual, 2GB: 8 hr (add about 2 hr for buildkernel)

(-j 2, -j 3, and -j 4 had problems for the G4's so I did not use -j for the G4's. portmaster -m MAKE_JOBS_UNSAFE=yes for the G4's for ports also seems to be frequently required and was involved in building xorg and xfce4.)

===
Mark Millard
markmi at dsl-only.net