Calibration uses a lot of internal variables, and it used to overflow
even a 3kB stack (by a little amount). This patch shares a temporary
buffer between two functions, so it takes 128 bytes less of stack, and
3kB fit for the wr-nic release.

The problem must still be studied, to reduce stack use.

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

This fixes a bug with negative hex numbers (that must
be printed as unsigned hex, without the minux sign).

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>

The rule for "git submodule update" was wrong, so the command was
always performed.  This is a pain if you are using commits in the
submodule that are not yet committed in the supermodule. This always
happens during development.

I'm sorry for the inconvienence, and I thank Aurelio for noting this.

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

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

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

We are always pretty near to the end of RAM, so now that we
have a reserved stack area, we can also check if it overlows.
When it happens, let's print a message forever, so a user who
comes and see what's up will see it eventually.

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

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

Pretty often it happens that during trials we pick wrong external
functions (e.g., "printf") and the final link claims horribly about
dozens of undefined symbols.  By having an intermediate wrc.o we can
easily check which are the symbols we really leave undefined for libc
and libgcc to fill, before they resolve them by asking for something
else.

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

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

This renames ram.ld to ram.ld.S, adding a preprocessing rule. The
commit has no techcnical effect: it is only preparation for the next one.

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

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

This adds copyright notes to all non-trivial source files,
unless they where already there (i.e. dev/endpoint.c alone).

I found authorship using the following script, run on commit "a2721762


documentation updated" (i.e, before automatic reindentation and other
trivial stuff by me):

  git grep -l . | grep -v sockitowm/ | \
  while read F; do
       echo "##### $F"
       git blame -w $F | \
         sed -e 's/^[^ ]* .//' -e 's/-[0-9][0-9]-[0-9][0-9] .*$//' | \
         sort | uniq -c | sort -rn
  done

Then I augmented each file with this boilerplate:

/*
 * This work is part of the White Rabbit project
 *
 * Copyright (C) 2011 CERN (www.cern.ch)
 * Copyright (C) 2011,2012 CERN (www.cern.ch)
 * Copyright (C) 2012 CERN (www.cern.ch)
 * Copyright (C) 2011 GSI (www.gsi.de)
 * Copyright (C) 2011,2012 GSI (www.gsi.de)
 * Copyright (C) 2012 GSI (www.gsi.de)
 * Author: Tomasz Wlostowski <tomasz.wlostowski@cern.ch>
 * Author: Grzegorz Daniluk <grzegorz.daniluk@cern.ch>
 * Author: Wesley W. Terpstra <w.terpstra@gsi.de>
 *
 * Released according to the GNU GPL, version 2 or any later version.
 */

Then I removed all the lines that didn't apply. Sometimes I reordered
the authors to reflect who is the main author.

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

Some individual files are LGPL, some are public domain,
some are GPL 2-only and most are GPL 2-or-later. You can
pick parts according to the respective license (marked
in the files themselves), but this one (GNU GPL version 2)
is what applies to the work as a whole.

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>

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

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

When we'll enable pp_printf, we'll need to resolve mprintf to
it. Local files are fixed by #define in wrc.h, but ptp-noposix has
explicit mprintf calls inside.  By using "PROVIDE" in the linker
script, we resolve mprintf to pp_printf only if no native mprintf
is part of the link stage. This has no effect if pp_printf is
not selected.

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

We include newlib headers (and link part of newlib), so any
locally-defined function must have the same prototype as standard
ones. We'll soon nee a "puts" implementation for pp_printf, so this
adds puts as an alias of uart_write_string (at no size cost) and makes
uart_write_string itself return integer (this costs a few bytes).

While I'm at it, change uart_write_byte to be slightly faster.

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

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

This comes from git://gitorious.org/rubi/pp-printf.git

 .
This commit just copies it over, without enabling it for compilation.

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>