I also turned to gui to void, as nobody checks the return value and
there's no error code to return in any case.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

When I add prototypes to headers, I ensure the source that
defines the function sees the header, to detect inconsistencies
(here, I add <wrc.h> to the includes).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This adds all missing prototypes and removes unused variables.
It includes a generic <wrc.h> (which btw includes the printf
prototype, not in uart.h any more) and other needed bits.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

There was a wrong "static" in a public header that triggered
my interest here. This fixes the warning and makes all functions
that are unused elsewhere really static. It also saves 50 bytes
in the final binary.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Remove all unneeded stuff from the linker script. Now the script only
includes what is actually used: it is easier to understand and change
(we'll add some local ELF sections), and it makes a better reference
for possible porting to a different architecture.

This commit also moves the default libraries from the linker script
itself to the LDFLAGS in Makefile, where they really belong.

CONFIG_DETERMINISTIC_BINARY confirms this makes no difference in
the generated binary (nor the ELF FWIW).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Instead of picking libraries from the linker script, common practice
is listing them in LDFLAGS. This simplifies review and modification.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

CONFIG_DETERMINISTIC_BINARY confirms this makes no difference in
the generated binary.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

It also used cflags-y in the same way.

CONFIG_DETERMINISTIC_BINARY confirms this makes no difference in
the generated binary, but it opens to more configuration options.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This also removes the run of $SIZE on all object files, as the
previous content of the screen is often more important for me.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

I prefer the comment separatos for sections to be different than those
for chapters (so I better see the structure while editing). It makes
no difference for others I'm sure.

Also, remove the info menus: they tend to rust in the source
while the good ones in the output are regenarated by ./fixinfo .
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

While compiling a noticed some overfull and underfull boxes, and
they are now fixed. Also, change the date and write that this is not
a release (while leaving references to the last release we had.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This doesn't change the behavior of any command, but adds revision.c
as a real source file rather than something created by make using
"echo ... > revision.c".  In any case, the file is recompiled
each time you call "make", so the date and time reflect the last
build.

However, this introduces CONFIG_DETERMINISTIC_BINARY (that nobody
sets at this point in time) to build a binary withouth __DATE__
and __TIME__ strings, so it is the same at every rebuild. See
next commit.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This has been verified to properly obey the new CONFIG_ETHERBONE
settin.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Add the lines from the .gitignore found in bst-kconfig. As in the
previous commit, use bst-kconfig commit v2.6.34-4-g8168c0b
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This undoes some differences between the upstream 2.6.34 kconfig and
the bst-kconfig out-of-tree port. This basically adds back
include/config/ and include/generated/ .
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This comes from the "bst-kconfig" project, which lives at
  git://git.informatik.uni-erlangen.de/vamos/bst-kconfig

In particular, I copied from commit v2.6.34-4-g8168c0b which exported
to a standalone package the kconfig code from version 2.6.34 of the
kernel. This commit copies the files literally, but it renames
"Makefile" to "Makefile.kconfig", as our Makefile is already in place.

It can be said that it is a lazy choice and the code is old, but I
also evaluated other projects as a source of Kconfig code.  In
particular, I studied busybox and buildroot; but the work needed to
extract the relevant code looks bigger. I also evaluated
kconfig-frontends, but it is not what we needed here (besides, it uses
automake and thus it's completely unmaintainable for me -- known
hackers say that autotools must die, and I share the idea).

If needed, we can replace this with a newer Kconfig setup later on.
At the time being, this is code base is more than enough for this
project.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Commit "5433fa2f Makefile: three fixes" introduced a bug, so that a
fresh clone fails to compile because of a missing "board.h" file.

This makes all objects depends on the symlink, which in turn is
removed by "make clean", so you don't need to start from a fresh clone
to experience the complete build.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Conflicts:

	Makefile

This fits in 80 columns and makes other space cleanups, but
there is no functional change.

We'll need to touch it seriously when adding Kconfig and ppsi as
an alternative to ptp-noposix, so I'd better have it readable before
changing the meat in there.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Conflicts:

	Makefile

I'd better set CROSS_COMPILE in the environment than change my path,
so CROSS_COMPILE is conditionally assigned.

The revision is better "git describe --dirty" than "git rev-parse". The output
is a short beautiful string like "wrpc-v2.0-13-g09a736d5"
with an optional "-dirty" trailer, to show we compiled a non-committed
change.

Finally, to execute commands we dont $(shell) at toplevel, but have
them in the build rule.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>