The commit 84d0689 "Nano-malloc: Fix for unwanted external heap
fragmentation" from newlib 4.1.0 introduced several optimizations,
one of which is as follows:
When the last chunk in the free list is smaller than requested,
nano_malloc() calls sbrk(0) to see if the heap's current head is
adjacent to this chunk, and if so it asks sbrk() to allocate the
difference in bytes only and expands the current chunk.
This doesn't work if the heap consists of non-contiguous regions.
sbrk(0) returns the the current region's head if the region has any
remaining capacity. But if this capacity is not enough for the second
(non-trivial) call to sbrk() described above, allocation will happen
from the next region if available. Expanding the current chunk won't
work and will result in a segmentation fault.
So this optimization needs to be reverted in order to bring back
compatibility with non-contiguous heaps. Before the next version
of newlib becomes available and gets updated in the GCC Arm Embedded
Toolchain, we work around this issue by including the fix in Mbed OS.
The linker prioritizes malloc() from the project to the one from the
toolchain.
Initially, every library greentea test has its project() creation in
their CMake. As running greentea using CTest move all greentea test
suite under one global project mbed-os and MBED_CONFIG_PATH set at
the root mbed os CMake under the condition BUILD_GREENTEA_TESTS
check so refactored platform greentea CMake accordingly.
The PSA Attestation test suite requires full RTOS. There is no need
to explicitly check PSA support, because when we build all greentea
tests from the top of Mbed OS, PSA tests only get included if PSA is
enabled.
Add an option to enable the greentea tests independently from the unit
tests.
We can't just use the typical BUILD_TESTING option to enable greentea
tests. BUILD_TESTING enables unit tests and fetches googletest, which
are compiled for the host. Greentea tests are cross compiled and require
a toolchain file. For this reason we add a new option just to enable
greentea tests, preventing build failures triggered by the unit tests
and googletest.
Individual libraries' `target_h` stub headers have now all been moved
from `mbed-headers-base` to `mbed-headers-<library>`.
Note: Even though headers previously in `target_h` are technically
stubs/fakes too, they are used by not only unit tests but also regular
libraries when compiled for unit tests, because no target-specific HAL
implementation exists in this case. In order for regular library
sources to pick up `target_h` headers, those headers must
* have the same names as regular headers
* appear first in include paths
This is why those headers are part of `mbed-headers-<library>` and not
`mbed-stubs-<library>`. Before this refactoring, `mbed-headers-base`
was the first in unit tests' include paths.
The stub header randLIB.h overrides the header of the same name in
platform/randlib/ which is an external repository vendored into
the mbed-os codebase. As the repository is synchronized regularly,
it is better not to put overrides there, so we put the randLIB.h
stub into the regular platform doubles directory.
Stubs previously in UNITTESTS/target_h/ have the same names as
regular Mbed OS headers, intending to override the latter directly.
We move platform target_h stubs into
platform/tests/UNITTESTS/doubles/platform/.
Note: nvic_wrapper.h is normally implemented and used by Mbed targets
as needed. But as unit tests do not have a real target, we treat it
as a stub for the platform.
ARMCC provides __semihost via compat header. As CMSIS 5.8.0 removed this compat header,
we need to explicitly include it to fix definition missing error.
Reference: CMSIS 5.8.0 known issues and https://github.com/ARM-software/CMSIS_5/issues/1211
This fixes the error about redefinition of enable/disable irq. we need compat header because of
semihosting (not yet provided in CMSIS).
Previously a test executable was recognised as a single test by CTest.
However, test executables usually contain multiple test cases, the
results of the test cases should be individually reported. With our
previous setup we could miss test case failures that didn't cause the
executable to return an error code.
This commit uses gtest_discover_test to discover all test cases in a
test executable. This enables CTest to match test passes and failures
from the googletest binary output.
The CircularBuffer doesn't allow pushing zero elements; you must push at
least one. Update the CircularBuffer unit test to avoid invalid use of
the CircularBuffer.
On a target that doesn't support Firmware Update, compilation still works, and any attempt to call the Firmware Update API returns a runtime error which is good enough.
When building greentea tests, each test is an executable with its
own output binary path. This is also the case when a user project
produces multiple executables. But the current implementation of
post-build operations always assumes there's only one executable,
at the root of the build directory.
The post-build command depends on Mbed target, and it always takes
the the executable we build as an input file. To achieve this, we
let each Mbed target (that has a post-build command) define a function
function(mbed_post_build_function target)
which takes a CMake executable target as an argument from which it can
get its binary path using generator expressions. It generates and adds
to the passed executable target a post-build custom command.
Notes:
* The function name needs to be exact, because CMake only supports
literal function calls - CMake can't dereference a function name from
a variable. To avoid multiple definitions of this function, each Mbed
target needs to guard it with a macro to check if the user is
building this Mbed target.
* `mbed_post_build_function()` is a function, but it is usually
defined by another macro rather than a parent function, because
nesting functions would make many variables inaccessible inside the
innermost `mbed_post_build_function()`.
* There's no more need to force regenerate images. Previously, post-
build commands were custom *targets* which always got to run, so we
force regenerated images on every build to avoid patching an image
that's already been patched once on previous build. Now post-build
commands are custom *commands* of the same executable target, and they
are only run if the executable target itself is rebuilt.
We have a central collection of "stub headers", which makes reasoning
about dependencies rather difficult, as it forces every stub library to
depend on all available stub headers. The standard approach would be for
each stub library to expose its public headers, and its dependents to
explicitly specify a dependency on the stub library containing the
headers it needs. This is a more modular design than creating a
header-only monolith library. Move the platform stub headers from this
central library into the mbed-stubs-platform library to increase
modularity.
mbed-stubs-connectivity now depends on the mbed-stubs-platform because
it requires some headers which were moved to mbed-stubs-platform.
Previously the platform stub library depended on `mbed-headers`, which
is a collection of all available headers in mbed-os. To make it easier
to separate the library, only depend on the headers we're actually using.
Move the header-only mbed-headers-platform library the unit test stubs
depend on into the platform component directory. This makes the platform
stubs more self contained and improves composition of the library.
Move the platform stub library into the platform component directory.
This change is so we can avoid duplicating the mbed-os source tree in a
central UNITTESTS folder.
subprocess.PIPE is used to enable the parent process to communicate with
the subprocess via pipes, which mean all stdout and stderr messages are
captured and returned as part of Popen.communicate's result tuple.
In our case, we want to display the error messages on the console, so we
don't need to capture the output from stdout.
Example of a typical error message before this change:
```
Traceback (most recent call last):
File "platform/FEATURE_EXPERIMENTAL_API/FEATURE_PSA/TARGET_TFM/TARGET_TFM_LATEST/scripts/generate_mbed_image.py", line 197, in <module>
sign_and_merge_tfm_bin(args.tfm_target, args.target_path, args.non_secure_bin, args.secure_bin)
File "platform/FEATURE_EXPERIMENTAL_API/FEATURE_PSA/TARGET_TFM/TARGET_TFM_LATEST/scripts/generate_mbed_image.py", line 81, in sign_and_merge_tfm_bin
" secure binary, Error code: " + str(retcode))
Exception: Unable to sign musca_b1 secure binary, Error code: 1
```
Example of the error message after this change:
```
Traceback (most recent call last):
File "/mbed-os/tools/psa/tfm/bin_utils/wrapper.py", line 13, in <module>
import click
ModuleNotFoundError: No module named 'click'
Traceback (most recent call last):
File "platform/FEATURE_EXPERIMENTAL_API/FEATURE_PSA/TARGET_TFM/TARGET_TFM_LATEST/scripts/generate_mbed_image.py", line 194, in <module>
sign_and_merge_tfm_bin(args.tfm_target, args.target_path, args.non_secure_bin, args.secure_bin)
File "platform/FEATURE_EXPERIMENTAL_API/FEATURE_PSA/TARGET_TFM/TARGET_TFM_LATEST/scripts/generate_mbed_image.py", line 80, in sign_and_merge_tfm_bin
raise Exception("Unable to sign " + target_name +
Exception: Unable to sign musca_b1 secure binary, Error code: 1
```
This is a significant improvement as now you can see what the reason for
the failure was.
Move /val and /pal directories into /test_abstraction_layers directory
and combine into one CMake target, mbed-psa-tal. Moved into seperate
directory in order to have own CMakeLists.txt, rather than adding to
/TARGET_MBED_PSA_SRV CMake file.
Assumption that greentea test file is always named main.cpp is
incorrect. Updated mbed_greentea_add_test() macro to make TEST_SOURCES
parameter compulsory, which is used to specify greentea test
file(s). This allows tests to use C, or have a different name.
Therefore also updated all pre-existing greentea test CMake files to
explicity add main.cpp to TEST_SOURCES.
Martin Kojtal
Lingkai Dong
Rajkumar Kanagaraj
Leon
mbedmain
Robert Walton
Jaeden Amero
Chun-Chieh Li
Jerome Coutant
Hari Limaye
Arto Kinnunen