- 28 Feb, 2013 16 commits
-
-
Alessandro Rubini authored
This somehow voids the effort of two commits ago. But we really don't need to save space when building a full exectutable linked with libc and all the rest. As a side effect, it allowed me to discover the wrong varargs to some messages (see previous commit). Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
printf with missing argument. Introduced in a8733f10, 52 commits ago. I didn't notice because I was not using CONFIG_PPSI_RUNTIME_VERBOSITY (as expected, the preprocessor hit again). Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
The pp_Vprintf (Verbose printf) macro, replaces PP_VPRINTF in the file to ease generation of dumps. All messages are build-time protected by "ifdef VERB_DUMP" anyways: by removing the dependency on CONFIG_PPSI_RUNTIME_VERBOSITY it's easier to enable the messages at build time because you just need to set USER_CFLAGS="-DVERB_DUMP" without the need to force RUNTIME_VERBOSITY as well. Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
The big dumping stanzas are already conditional under if(pp_verbose_dump), so they only exists if VERB_DUMP or VERB_LOG_MSGS is defined at build time. This commit adds the same build-time protection to other messages related to packet dumps (all of them rely on a build-time protected subfunction anyways). Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
The three "raw" fields claim to exist for debugging, but since we have several TimeInternal structures around they cost quite some space. The fields were only copied to TimeInternal from network frames, and there is not explicit code to dump them. Thus, this commit removes them but leaves the code in place, so it can be re-enabled if the need arises. (I use the ugly "#if 0"). Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
sometimes I jumped to "no_incoming_msg" in case of error. This renames the labels to "out", as the catch both error and normal exit paths. Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
set_TimeInternal was partial anyways, since TimeInternal is more than seconds and nanoseconds (even if only WR uses it). It was only used to zero the values, so use memset instead, within the new inline clear_TimeInternal() Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
This stuff must be reviewed again anyways, because I'm not sure at all about what these numbers add (I can't find proper docs or examples). Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
- 27 Feb, 2013 24 commits
-
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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>
-
Alessandro Rubini authored
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).
-
Alessandro Rubini authored
Signed-off-by: Alessandro Rubini <rubini@gnudd.com>
-