[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Xorg fails on ATI after update (mmio aperture)



On 03/06/2017 11:04 AM, Riccardo Mottola wrote:
> Exactly, that "proof" means that the changes comes from the kernel or
> that the driver needs updating to work with the new kernel.

Or the kernel needs to be fixed in case it's a regression.

> Compiling kernels is a bit tight on this iBook. What road can i follow
> to help you track down changes? Can I get easily a dpkg package of intermediate kernel
> versions to bisect things?

You can just cross-compile the kernel from a fast amd64 machine. If you
are on Stretch or Unstable, you can install a cross-compiler for PowerPC
with:

	# apt install gcc-6-powerpc-linux-gnu

and you will most likely also need things like the ncurses development
headers if you want to use things like "make menuconfig". So, I'd just
install the build dependencies for the Linux kernel:

	# apt build-dep --arch-only linux

> Are there changes in the kernel configuration itself? We go from 4.7.0-1 to 4.9.0-2

Check for changes here:

	> https://anonscm.debian.org/cgit/kernel/linux.git/tree/debian/config

and in the Debian changelog:

	> http://metadata.ftp-master.debian.org/changelogs/main/l/linux/unstable_changelog

> If you can point me to an intermediate version I'd like to try it!

You will find these versions on:

	> http://snapshot.debian.org/

Follow the instructions on the webpage on how to add a sources list entry for
a certain snapshot date. Then run "apt update" with verification disabled
and install the kernel version of your choice.

> Not all acceleration, but most of it - that is the best I could get about a year ago withotu having very strange display issues.
> I remember this was quite a still  bette than "accel off" or fbdev which is quite horribly slow. I'd love to nable some of it again and gain again some speed
> which was usable about 2 years ago.
> 
> Of course the goal is to get most features and acceleration working (again),

Then you need to provide as many datapoints as possible. Trying to bisect the
kernel version which broke the driver is already the right idea.

I would start working with snapshot.debian.org to find the Debian kernel
version which broke the driver. Then we can inspect the changes for the
first bad version and possibly continue with "git bisect" if we cannot find
any obvious change.

Adrian

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaubitz@debian.org
`. `'   Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913


Reply to: