New satconfig version (for 4.3.0)

[Vegard Nossum]

Hi all,

I've rebased my satconfig branch onto v4.3 and fixed a couple of bugs
which now actually makes satconfig quite usable (depending on what you
feed it as input, of course).

Known bugs/TODO:

1. Actually call conf_write_autoconf() ourselves and figure out why it
changes a lot of config options compared to not calling it + running
yes "" | make silentoldconfig

2. Write a regression/test suite using isolated Kconfig fragments to
test individual features of the SAT translation (e.g. prompts,
defaults, choices, dependencies, etc.)

3. Figure out why yes "" | make silentoldconfig changes anything at all

In the meantime, if you want to give it a try, here's an example that
gets a fairly minimal UML kernel running:

git fetch g...@github.com:vegard/linux-2.6.git v4.3+kconfig-sat
git checkout FETCH_HEAD

cat > .satconfig <<EOF
CONFIG_MODULES=n

# standard features
CONFIG_PRINTK=y
CONFIG_BINFMT_ELF=y
CONFIG_BINFMT_SCRIPT=y

# uml-specific
CONFIG_STDERR_CONSOLE=y
CONFIG_HOSTFS=y
CONFIG_BLK_DEV_MD=y
EOF

# Run the solver with above .satconfig file as input
make ARCH=um satconfig

# Build the kernel
make ARCH=um -j8

# Run it
./vmlinux rootfstype=hostfs rw init=/bin/bash

Let me know if you have any questions/comments or run into any problems.


Vegard

[Luis R. Rodriguez]

On Sun, Dec 27, 2015 at 06:54:27PM +0100, Vegard Nossum wrote:
> Hi all,
>
> I've rebased my satconfig branch onto v4.3 and fixed a couple of bugs
> which now actually makes satconfig quite usable (depending on what you
> feed it as input, of course).

Nice work! My recommendation is to just always work on Linus'
latest tree, and rebase regularly so you don't have to address
large chunks of code shuffling.

Luis