This introduces "-DDIAG_ERR" at build time and pp_error() as a
conditionally-compiled and conditionally-enabled thing.  I'm not sure
error messages should be missing from the binary by default, but I'll
evaluate better at a later time, after debugging my own errors.
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>

Now you can set VERB_DUMP, VERB_SERVO, VERB_FRAMES, VERB_TIME
separately.  For example

    make USER_CFLAGS="-DCONFIG_PPSI_RUNTIME_VERBOSITY -DVERB_SERVO"
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This has currently no effect, but later I'll have an option to enable
verbose messages for individual contexts (message dumps, servo
operations, frame send/receive, time operations).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This only removes the hairy ifdef, to convert them to normal if
functions.  There is no effect on the generated binary, but it's now
less easy for me to introduce errors.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This commit makes no technical changesin the output, but it creates
<ppsi/diag-macros.h> (always includes) to split out all the
diag-related stuff.  The macros in there are the same as in previous
commits, but they are documented.

This also introduces pp_verbose_messages (which is 1 or 0) to allow
the various "#ifdef VERB_LOG_MSGS" to become language if() instead.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

There is no way to set VERB_LOG_MSGS. This Makefile change allows
to set USER_CFLAGS, for example like this:

   make USER_CFLAGS="-DVERB_LOG_MSGS -DCONFIG_PPSI_RUNTIME_VERBOSITY"

and then use "-V -V" to enable all messages.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

pp_printf calls puts, but that's the "embedded" puts, the one that
doens't add a trailing newline. So to avoid extra lines and split
messages in the diagnostic output we need to link the fake puts, that
calls pp_puts instead (in turn, pp_puts for arch-gnu-linux relies on
fputs, which doesn't add the trailing newline).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

The time operations are get, set and adjust. They are methods of a
single structure in order to simplify use of custom timers (and make
them separate works according to the LGPL).

This patch set introduces specific <arch>-time.c files where all the
timing stuff lives (not much, actually). Sometimes this turns
<arch>-io.c into an almost-empty file, but that how history changes
stuff.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Nobody was implementing this timer_adjust_all, and I'm going to
rewrite the timer management anyways, because it is a horrible
complication just to manage some timeouts for packet transmission.

So let it RIP, and let's happily leave without it.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

Instead of having several external functions for networking, only
have one symbol: the structure of network operations. This simplifies
stuff a while, and allows more static symbols and better encapsulation.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This bug was not exposed, as our extension offers both the open and
close hook. Bug introduced in 9e4a90c7, 20 commits ago.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

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

I confess I repent having started this. And some of the files are really
in a terrible status and need rewrie anyways (I'm not pointing fingers,
I know the history of the project and I'm aware this is what I deserve).
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 unifies PP_ADJ_FREQ_MAX, that was defined in several places with the
same value.

Unfortunately, it isn't clear at all to me what is this value (looks
like arch-spec uses it as nanoseconds and not as an adjustment).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This kills an ifdef that was due to die anyways. The only disadvantage
is that bare-* architectures can't be passed arguments to force
slave-only (while arch-gnu-linux can received it on the command line).
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This patch set removes duplicate code in ext-whiterabbit, using hooks
in proto-standard to call specific extension code.

There is still some duplication in wr-msg.c from msg.c), so I'll need
to touch stuff again, but this is good and stable enough to merge.

Unfortunately, the set of extension hooks is specifically tailored to
white rabbit, and we'll need to make it more orthogonal and solid as
soon as another extension must be supported (or as soon as we have
time).

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

The ifdef was used to print smaller messages if the daemon is known
to be slave only. Since this PPSI_SLAVE is going to die, remove the
ifdef and keep the more complete branch of the code.
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>

Since timeInternal is seconds and nanoseconds, avoid copying both
and a temporary variable by just storing them to the target structure.

This is also meant to avoid a WR-specific hook in the next commit,
since timeInternal in WR is more fields than seconds and nanoseconds.
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 removes the WR-specific state-initializing.c by introducing
the init() hook for the standard state-initializing.

Meanwhile, clean up error management a little in the init state.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

This new "struct pp_ext_hooks" is used to include pointers to
extension-specific hooks that allow to avoid code duplication.
I'm now going to apply it to ext-whiterabbit, one state at a time.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>

struct DSPort now has a void pointer called ext_dsport that is used by
each extension to store its own local data.  White Rabbit, thus,
is not a special case any more in this respect.
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>