ticker_insert_event() can crash on KLXX (and probably other platforms) if an event is inserted with a timestamp before the current real time.
The problem is easy to trigger: you just need to set up a Ticker object, and then disable interrupts for slightly longer than the Ticker object's interval. It's generally bad practice to disable interrupts for too long, but there are some cases where it's unavoidable, and anyway it would be better for the core library function not to crash. The case where I had an unavoidably long interrupts-off interval was writing flash with the FTFA. The FTFA hardware prohibits flash reads while an FTFA command is in progress, so interrupts must be disabled for the whole duration of each command to ensure that there are no instruction fetches from flash-resident ISRs in the course of the execution. An FTFA "erase sector" command takes a fairly long time (milliseconds), and I have a fairly high frequency Ticker (1ms).
The problem and the fix are pretty straightforward. ticker_insert_event() searches the linked list to figure out where to insert the new event, looking for a spot earlier than any event currently queued. If the event is in the past, it'll usually end up at the head of the list. When the routine sees that the new event belongs at the head of the list, it calls data->interface->set_interrupt() to schedule the interrupt for the event, since it's the new soonest event. The KLXX version of us_ticker_set_interrupt() then looks to see if the event is in the past, which we've stipulated that it is, so rather than actually setting the interrupt, it simply calls the handler directly. The first thing the Ticker interrupt handler does is re-schedule itself, so we re-enter ticker_insert_event() at this point. This is where the problem comes in: we didn't finish updating the linked list before we called set_interrupt() and thus before we recursed back into ticker_insert_event(). We set the head of the list to the new event but we didn't set the new event's 'next' pointer.
The fix is simply to finish updating the list before we call set_interrupt(), which we can do by moving the obj->next initialization ahead of the head pointer update.
Add an 'all' parameter to get_flash_algorthim_binary. When set
an enumerator of all flash algorithms is returned. rather than just
the first flash algorithm.
In targets.json, add lf_clock_rc_calib_timer_interval and
lf_clock_rc_calib_mode_config setting which are essential when using BLE
In mbed_overrides.c, create valuables instead of doing printf, those
valuables are intended to be used for debugging in runtime.
I accidentally broke only uARM mbed 2 builds. Here is the story: When
scanning for resources, toolchains look for any `TOOLCHAIN_<classname>`
folders to include. These `<classname>`s mostly match the name passed in
on the command line with one exception: `uARM` on the command line maps
to `ARM_MICRO` the class. This would not be a problem except for the bug
that I introduced in a prior commit. The bug is that the mbed2 library
builds will use the name passed in on the command line to construct
`TOOLCHAIN_<cli-name>`. This will not match when scanning.
I fixed it by translating the `<cli-name>` into the `<classname>`.
Currently, there is the issue in register setting into serial_format().
The issue is that parameter for baudrate setting is initialed zero in this function.
In baudrate is less 9600bps, the issue occurs. In baudrate is over 9600, not occurs. When call serial_baud() before serial_format(), baudrate override and set with an incorrect value.
Therefore, I fixed register setting not to set the parameter of baudrate setting in serial_format().
The `sleep` function as been changed into `hal_sleep` by #3607.
Unfortunately the call to `sleep` in the hal_patch for the NRF51822 has not been
updated to `hal_sleep`. The result was a link time error for targets based on
NRF51822_LEGACY compiling with the mbed OS 5 tree.
The 'TARGET_M%' rules depend on files generated by the 'rsync' target,
so make this dependency explicit by setting 'rsync' as a pre-requisite
of 'TARGET_M%'. The 'rsync' dependency was removed from 'publish' to
avoid the case where make would select one of the 'TARGET' rules before
completing 'rsync', which would cause the build to fail.
Additionally, also moved the core libs selection in the 'TARGET_M%'
pattern rule from the pre-requisites into the rule's recipe. This is
required because when the wildcard expression used before as a
pre-requisite is expanded (make's 1st phase), it won't find any files
(not built yet via 'rsync' target), so it won't create the associated
'TARGET_M%' rules, finally causing the 'publish' target building to fail
due to missing rules for 'TARGET_M3' and 'TARGET_M4'.
With this change and the previous one, it is ensured that 'rsync' is
done before executing the recipe for 'TARGET_M%', so the required core
libs are already available and can be used from withing the recipe. The
same wildcard pattern is used as before.
The issues being fixed are visible with GNU Make 4.2.1, but not seen
with GNU Make 3.81.
Fixes issue #3905.
An earlier patch in this series changed the API for
`build_api.prepare_toolchain`. This commit updates the `find_test`
function to call `prepare_toolchain` correctly.
An earlier patch in this series relies on the assumption that all
toolchain construction goes through `prepare_toolchain`. This is still
not the case. The only remaining user of the `mbedToolchain` object that
does not go through `prepare_toolchain` is the static analysis scanner.
It's basically dead-code at this point. I say we remove it. So this
patch removes it.
The prior patch in this series makes the assumption that any building
will go through `build_api.prepare_toolchain`. This was not a valid
assumption for the mbed2 build process. So, instead of maintaining 2
ways of using the toolchain classes, I elected to unify on
`prepare_toolchain`.
If the free memory on a device is small enough then the calculation to
determine heap size could underflow to a large value. If this happens
then malloc will never return failure and instead will cause a crash on
allocation. This patch prevents the underflow so malloc works as
expected even with low amounts of free memory.
This is a bug fix for the following bug (Github issue #437):
If two builds were run specifying a non-default build folder, the second
build would fail to link with duplicate symbols and may not fit on the
device. The root of this problem is that these non-default build folders
are not ignored by scan-resources, and therefore included in the build.
We fix this bug by ignoring the build directory passed into the tools.
mjrgh
Anna Bridge
Russ Butler
TsungtaWu
Sam Grove
Jimmy Brisson
Christopher Haster
Radhika Raghavendran
TomoYamanaka
Mohammad Azim Khan
Martin Kojtal
Simon Vogl
adustm
Vincent Coubard
Andres AG
Ricardo Silva
Laurent MEUNIER