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

You can pass architectures (for example "./MAKEALL spec"),
specific settings (for example "./MAKEALL CONFIG_NO_PRINTF=y")
or "-u" to ask the printout of undefined symbols in each ppsi.o
being built.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

If you "export CONFIG_NO_PRINTF=y", no printf will be built in the
binary. This is needed when ppsi is built within wrpc-sw (which has
its own pp_printf implementation).

This will become a Kconfig item, when Kconfig is added.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

Now that we have vprintf, we can save a buffer in the stack and
simplify pp_diag_printf and pp_timed_printf.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

This patch-set removes the old diag/printf.c and related files, to use
the "official" pp_printf, which is now published separately and
included here in its entirety with not changes from its upstream repo.

Please note that HAS_DIAG and HAS_FULL_DIAG is not supported any more.

While we have no Kconfig support (coming soon), please see ./MAKEALL
to find how to pass verbosity (and thus size) options to pp_printf.

All commits build for all architectures.

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 is a copy of pp_printf, that originated from ppsi's diagnostic
code and is now a separate package, with a few more features.
The upstream repository is git://gitorious.org/rubi/pp-printf.git .
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This patch-set removes most of the awful pp_ libc replacements (like
pp_memcpy) that I used in the initial ptp-proposal project.  I now
think using standard names is better and easier. The build for all
architectures has no conflicts with this approah (all commits build).

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>

Even the freestanding lm32 compiler has <string.h> (because gcc builds
newlib by default as a minimal fallback).  Since I want to get rid of
pp_strlen and such stuff, we'd better include <string.h>, win size_t
and use the standard definitions for standard functions.

When we port to systems with no <string.h> we can provide it inside
its own arch- directory. So no harm is done and the code is simplified.
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 file is meant to share libc functions among non-libc-using
architectures.  In this commit it still exports the pp_* version,
which are removed later in this patch-set.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

If the user has WRPCSW_ROOT in the environment, build for arch-spec
as well, because we can refer to wrpc-sw for drivers.  When building
for spec, build the whiterabbit extension as well.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Trivial changes: exit code should not be 0 on error, we should exit
where we say so to the user and give more informative context.
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>

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

This is the verbatim copy of what I wrote as example code for an
article about time stamping. It's called "mini time protocol".  Some
of the code, in turn, comes from my previous experiments with White
Rabbit.

I'd better have the programs here as I use them often. As time permits
I'll add some more features and documentation in the ppsi manual.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This program monitors clock_monotonic and clock_realtime every 10ms
(or a different time lapse passed on the command line).

Whenever the difference between them is more than 0.5ms apart from the
previously-reported difference, it prints the current difference.  All
such differences are relative to the initial offset of the two clocks,
when the program started. The difference is prefixed with the current
date and time, to help logging.

This is something I already had in my toolbox, so I could see the leap
second that occurred on Jun 30th 2012.

Example:
   tornado.root# ./tools/chktime &
   [1] 23761
   ./tools/chktime: looping every 10 millisecs

   tornado.root# ./tools/jmptime .003
   Requesting time-jump: 0.003000 seconds

      13-02-11-12:22:44:       2997 us

   tornado.root# ./tools/jmptime .003
   Requesting time-jump: 0.003000 seconds

      13-02-11-12:22:52:       5994 us

   tornado.root# ./tools/jmptime -.006
   Requesting time-jump: -0.006000 seconds

      13-02-11-12:22:58:        -10 us
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Like adjtime (previous commit), this is used to adjust the local time,
but in this case the program performs a time jump (warp) using
settimeofday().
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This simple tool requests a time adjustment, specified in seconds
expressed as a floating point values ("0.001", "-.2", "-4" ...).

It can be used to force time differences between hosts while checking
how ppsi (or other time daemons) work.
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>