The current error messages are a bit cryptic, rework them to make them
more meaningful:
- add an extra message on the first error to explain what the errors
refer to.
- rework the error message to be more explicit.
- rework the priority string print to use a LEVEL+offset format to
somehow highlight that the number is the offset from the level, not
the actual priority.
- print the init function name in addition to the devicetree path.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Since bb590b5b6e introduced ordinals in the priority sequence, the "same
priority" case cannot happen anymore, furthermore the priority value in
the script is now the position of the function in the init sequence, so
if two devices have the same priority there's something real bad going
on.
Drop all the "same priority" handling code and tests, convert the case
into ane exception instead. Drop the init stubs as well from the test,
they are not required anymore.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Now that child nodes are handled by edtlib there's no need to parse
child nodes in check_init_priorities anymore.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Rework check_init_priorities to use the main executable file instead of
the individual object files for discovering the devices.
This should make the script more robust in case of stale files in the
build directory, and also makes it work with LTO object files.
Additionally, keep track of the detected init calls, and add a handy
"-i" option to produce a human readable print of the initcalls in the
call sequence, which can be useful for debugging initialization problems
due to odd SYS_INIT and DEVICE interactions.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Log the device priority on the file debug output print, this is
convenient to list out all the defined devices and their priority, even
if they don't depend on each other.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Update the script to parse the new section naming. The ordering type
is converted from an integer to a tuple, which still compares correctly
due to the elementwise behaviour of tuple comparison.
Signed-off-by: Jordan Yates <jordan.yates@data61.csiro.au>
The new stack zephyr,cdc-acm-uart driver has two separate init path, one
of which kicks in before the USB stack to start buffering log messages.
This generates a false positive in the build time priority checking,
adding a check to ignore that compatible entirely.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Add support for more verbose output enabling debugging output and add a
first debug print to output the file to device path mapping. Found this
to be extremely useful to find what file is instantiating a specific
device for non obvious cases.
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
Add a check_init_priorities.py scripts. This goes through all the object
files looking for known initialization sections and building a map of
device ordinals and their effective initialization priority. Then
compares that with the list of dependencies generated based on the
device tree and reports any warning (dependent devices, same priority)
or error (incorrect priority).
Signed-off-by: Fabio Baltieri <fabiobaltieri@google.com>
