We don't yet support all combinations of configurations. With all.sh as it
currently stands, component_test_psa_crypto_config_accel_cipher results in a
build where PSA_WANT_ALG_CMAC is disabled but CMAC operations succeed
nonetheless, going via the driver. With the systematic testing of
not-supported operations, this now results in a test failure.
The code in all.sh does not respect the principle documented in
df885c052c701b89c24ad5f305114f8f8b39d750:
> The PSA_WANT_* macros have to be the same as the ones
> used to build the Mbed TLS library the test driver
> library is supposed to be linked to as the PSA_WANT_*
> macros are used in the definition of structures and
> macros that are shared by the PSA crypto core,
> Mbed TLS drivers and the driver test library.
Disable PSA_WANT_ALG_CMAC completely in this test component. This is not
wrong and it makes the test component pass.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
To be able to test utility programs for an absence of time.h, we need a
baremetal config that is not crypto only. Add one.
Signed-off-by: Daniel Axtens <dja@axtens.net>
baremetal compiles should not include time.h, as MBEDTLS_HAVE_TIME is
undefined. To test this, provide an overriding include directory that
has a time.h which throws a meaningful error if included.
Signed-off-by: Daniel Axtens <dja@axtens.net>
In build_arm_none_eabi_gcc_m0plus, use baremetal_size instead of baremetal
as the configuration, i.e. exclude debugging features. This job is the only
one switching to baremetal_size because it's our primary point of reference
for code size evolution, and which is the only job where we display the code
size built with -Os so it's presumably the only job for which we really care
about a meaningful code size report.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
The race condition mentioned in the previous commit
"Stop CMake out of source tests running on 16.04"
has also been observed with test_cmake_as_subdirectory on 3.1 and can
presumably happen on 2.28 as well. So skip it on Ubuntu 16.04 as well.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Running the out of source CMake test on Ubuntu 16.04 using more than one
processor (as the CI does) can create a race condition whereby the build
fails to see a generated file, despite that file actually having been
generated. This problem appears to go away with 18.04 or newer, so make
the out of source tests not supported on Ubuntu 16.04
Signed-off-by: Paul Elliott <paul.elliott@arm.com>
Add ECDSA and RSA signature acceleration testing
with signature capabilitites removed from the
Mbed TLS library.
Signed-off-by: Ronald Cron <ronald.cron@arm.com>
This commit removes the test_psa_crypto_config_basic
all.sh component that can no longer work without
adapting it to the separately compiled test driver
library. This component is replaced by several
components in the following commits to test various
type of acceleration independently.
Signed-off-by: Ronald Cron <ronald.cron@arm.com>
Non-regression for the fix in https://github.com/ARMmbed/mbedtls/pull/5126:
libmbedtls and libmbedx509 did not declare their dependencies on libmbedx509
and libmbedcrypto when built with make.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Set the build type to Release (-O2) when running CPU-intensive tests (ssl-opt,
or unit tests with debug features). A build type of Check (-Os) would be best
when the main objective of the build is to check for build errors or warnings
and there aren't many tests to run; in this commit there are no such test
cases to change. Only use cmake with no build type (which results in not
passing a -O option, and thus missing some GCC warnings) when exercising cmake
features.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Use Release mode (-O2) for component_test_full_cmake_clang which runs SSL
tests.
To have some coverage with Check mode (which enables more compiler warnings
but compiles with -Os), change a few other builds that only run unit tests
at most to Check mode.
Don't add any new builds, to keep the total build volume down. We don't need
extensive coverage of all combinations, just a reasonable set.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
SSL testing benefits from faster executables, so use -O2 rather than -O1.
Some builds use -O1, but that's intended for jobs that only run unit tests,
where the build takes longer than the tests.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Otherwise $COMMAND_LINE_COMPONENTS would try to expand wildcard patterns
based on files in the current directory.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Have simpler patterns related to 'test' (the central objective being to keep
going if 'make test' or 'tests/...' fails, but not if 'make tests' fails).
Add 'cd' as a can't-keep-going command.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Count invocations from 1 to n instead of n to 1.
Explain how changing the loop variable would cause an error if the function
was not executed in a subshell.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
all.sh restores **/Makefile from git in case the version in the worktree was
from doing a cmake in-tree build. Instead of doing this unconditionally, do
it only if the toplevel Makefile seems to have been automatically
generated (by cmake or otherwise, e.g. by mbedtls-prepare-build). This way
all.sh no longer silently wipes changes made to Makefile but not committed yet.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Back up the config files at the beginning of all.sh, rather than before each
component. In particular, create the backup before running cleanup for the
first time. This fixes#3139 (all.sh using a config.h.bak from a previous
job), and makes all.sh more robust against accidentally using a modified
config.h midway through because a component messed with the backup.
Use a different extension (*.all.bak rather than *.bak) for the backups.
This is necessary to ensure that auxiliary scripts such as depends*.pl that
make their own backup don't remove all.sh's backup, which the code from this
commit does not support.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Generate programs/test/cpp_dummy_build.cpp dynamically instead of
maintaining it manually. This removes the need to update it when the list of
headers changes.
Include all the headers unconditionally except for the ones that cannot be
included directly.
Support this dynamic generation both with make and with cmake.
Adapt all.sh accordingly. Remove the redundant C build from
component_build_default_make_gcc_and_cxx (it was also done in
component_test_default_out_of_box), leaving a component_test_make_cxx. Also
run the C++ program, because why not. Do this in the full configuration
which may catch a bit more problems in headers.
Fixes#2570 for good.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Signed-off-by: Thomas Daubney <>
Don't default to unbridled -j, which causes a load spike and isn't really
faster.
"Number of CPUs" is implemented here as a reasonable compromise between
portability, correctness and simplicity. This is just a default that can be
overridden by setting MAKEFLAGS in the environment.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
redirect_out was no longer used and redirect_err was only used to
quiet dd. Change the dd invocation to only print diagnostics on
error (on platforms where this is possible).
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This is necessary to actually keep going and finish the
component-specific cleanup in component_test_cmake_out_of_source if
ssl-opt.err is non-empty.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
In all.sh, when an explicit list of components is specified, error out
if one of the components is not known or not supported. Patterns that
happen to match zero components are still effectively ignored.
Fix#2783
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
