Page 4 of 30
Posted: Mon 02 Feb 2015, 07:32
by stemsee
Ah!
I never tested that scenario! The config file is configured for features that are not in the old kernel version, so what happens? Will need to enter source dir and examine .config file. Maybe in this case there needs to be a manual 'make menuconfig' interaction. An option I was intending to implement anyway! I think that when one enables IA32 there appear two or three more options to configure. When it is not set then those lines are no longer in the config file. I should be able to hack it one way or another or even include 3 alternative config files for such a scenario, for old configs, maybe Fd32 and Fd64 configs from kernel kit!
Let me have a think.
Re: no-ia32
Posted: Mon 02 Feb 2015, 18:03
by stemsee
Dry Falls wrote: Slapping it in externally is no good because the kernel and modules (builtin, etc.) have to agree.
I think you can use the same config and kernel source and manually build said modules shoud be fine!
http://sourceforge.net/p/tianocore/edk2 ... /build.bat
Posted: Mon 02 Feb 2015, 18:06
by stemsee
Could you send me the .config file from inside the sources?
configs
Posted: Mon 02 Feb 2015, 20:33
by Dry Falls
not sure if the attachment worked.
Posted: Mon 02 Feb 2015, 21:54
by stemsee
Here they are edited.
after copying them back to the sources dir open a terminal in there and
1. make oldconfig = recompile entirely
Posted: Tue 03 Feb 2015, 00:04
by Dry Falls
build.bat didn't work even after removing the dos language (rem).
Got the sources fixed but am having trouble rebuilding the modules (remember, I'm totally new at this).
first three lines of xerrors log:
Code: Select all
Window Manager: CURRENTWM=openbox-session
/usr/bin/start32: line 35: /usr/bin/32/bash: cannot execute binary file
cups: unable to restart scheduler.
I tried running the build script again and it finds the kernel tar.xz but not aufs...tar.bz2 (I slipped them into vanilla). removed them and tried a new download but in both cases got
or some such.
getting confused, will try a gin and tonic with three aspirin.
df
Posted: Tue 03 Feb 2015, 10:13
by stemsee
EDIT: Use my script. Then stop it 'ctrl+c' after it says kernel is configured. This must be done to get the kernel sources patched with aufs etc. Then enter the kernel sources dir. Copy the configs i sent in to there. Then open teerminal and type
Posted: Tue 03 Feb 2015, 21:46
by stemsee
Here is Stems-Unattended-Kernel-Kit SUKK-0.1
Added option to manually configure kernel.
Change aufsv to match kernel_version for all configs.
Also backs up edited .config from sources with date stamp per kernel compiled.
Fixed logic bugs and tested thoroughly.
Creates tar.bzip2 kernel package for use in woof-ce next.
If no bug is found I pretty much feel this is the release candidate.
Lacks only a gui perhaps.
Posted: Wed 04 Feb 2015, 06:11
by stemsee
experimental
Added code to build authentic debian kernel packages!
Posted: Thu 05 Feb 2015, 13:11
by stemsee
Building the authentic debian packages requires fakeroot and kernel-package to be installed, however there seems to be no way to test both packages, especially on a deb puppy. Also the old method has a replacement.
old
Code: Select all
Patches
make-kpg clean
make menuconfig
time fakeroot make-kpkg -j4 --initrd kernel_image kernel_headers
new
Code: Select all
Patches
make-kpg clean
make menuconfig
make deb-pkg
Didn't work for me on puppy or makululinux.
New sukk-0.3 changed the logic entirely. Old complicated out - new simple in.
Now the three builds are three functions which can be called in one of seven orders; 1, 2, 3,
12, 13, 23, or
123.
Posted: Fri 06 Feb 2015, 00:15
by stemsee
A few bugs ironed out:
1) importing variables from build.conf s (kept building 64bit)
2) stop it creating new dist/packages when finished compiling.
3) couldn't find aufs package for build 2 & 3
4) when building only 2 or 3 couldn't find kernel source package.
5) Wrapped test around make clean and make-kpg clean.
6) removed duplicate code for unpacking aufs.
7) reduced multiple instances of mkkernsrc to one.
Posted: Fri 06 Feb 2015, 18:20
by stemsee
Here is the final script.
Next to add a gui.
Posted: Sun 08 Feb 2015, 23:40
by stemsee
Just when you thought ... !
I left out the patches in 0.5
Added make oldconfig option default. Which ever options is used for the first compile will be carried to the rest.
Posted: Mon 09 Feb 2015, 07:20
by stemsee
Back on track. Compiled 3.18.6-64 and pae ... running on it now,.
Posted: Tue 10 Feb 2015, 12:17
by Gobbi
Hi @stemsee ,
Using sukk-0.6 on FD630 - only the '64' . Attached the log . IF it would be a good try - which are the first steps to do to go on and use the new kernel ?
Posted: Tue 10 Feb 2015, 19:48
by Gobbi
Following
@rufwoof's idea - it cross my mind to open the Fatdog's intrid , replace kernel-modules.sfs with kernel-modules-3.18.6-Fatdog.sfs from /sukk-0.6/dist/packages64 and then renamed it . Repacked the intrid , threw it together with the new vmlinuz into a frugal install and here I am : I'm posting from it .
Of course no fglrx , but it's really cool .
Nice job ,
@stemsee 
Posted: Tue 10 Feb 2015, 19:50
by stemsee
In fatdog click on initrd. It will open. Replace kernel-modules.sfs click on repack-invited script. Copy vmlinuz to boot folder and reboot PC.
Ahh ok! I couldn't,t compile on fatdog or lighthouse!!!???[/list]
edit I wrote this response on an android tablet ... auto complete!
kernels 3.18.6 pae 64
Posted: Wed 11 Feb 2015, 10:59
by stemsee
Posted: Sat 14 Feb 2015, 12:02
by stemsee
updated configs, especially pae.
Posted: Sat 14 Feb 2015, 14:24
by rufwoof
Gobbi wrote:Following
@rufwoof's idea - it cross my mind to open the Fatdog's intrid , replace kernel-modules.sfs with kernel-modules-3.18.6-Fatdog.sfs from /sukk-0.6/dist/packages64 and then renamed it . Repacked the intrid , threw it together with the new vmlinuz into a frugal install and here I am : I'm posting from it .
Of course no fglrx , but it's really cool .
Nice job ,
@stemsee
Great stuff Gobbi - and big thanks to stemsee
I've just extracted my initrd which contains my Slacko 5.3.3 pup sfs, unsquashed that pup sfs and dropped in EmSee's firmware and modules as replacements into /lib, copied the DOTconfig for the 32 PAE into /etc/modules .... and then copied all of those extracted folder/file contents directly over the normal (small) initrd contents ... i.e. to form a virtual full install - but that boots/resides in memory. Edited /init to just exec /sbin/init and nothing else (so no puppy layered file system setup up etc).
Reformed the intird using lz4 compression (with -l (little L)) switch, dropped in EmSee's vmlinuz and grub4dos booted using those vmlinuz and newly formed initrd.lz4 files.
All seems to be running great.
I mounted my external LibreOffice sfs, and cp -rs /mnt/lib/* /. symbolic linked copied the contents to the root directory and Libre's running fine.
Downloaded the latest firefox and that's ruhning ok as well.
I did try creating a lz4 squashfs - but the kernel doesn't recognise/load that. Looking at the config files (which I know very little about), I suspect the kernel has the squashfs suppoft of lz4 as a needed to be loaded module (outside of my area of knowledge).
So looks like the 64 bit and 32 bit PAE are both working ok.