Add gitlint configuration to allow contributor to check its
commit message compliance before sending its pull request.
The configuration files has been copied from Zephyr project.
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
Waiting the wiki, add a chapter to inform how to contribute.
this chapter is inspired from zephyr documentation.
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
Align coding rule with checkpatch recommendation concerning
the comparison to null.
Some comparisons in macro have not be fixed for readability.
Examples of code not updated:
VQ_PARAM_CHK(ring == NULL, status, ERROR_VQUEUE_INVLD_PARAM);
VQASSERT(vq, cookie != NULL, "enqueuing with no cookie");
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
All that the zephyr_library_named(foo) function guarantees is that
after calling it, the variable ${ZEPHYR_CURRENT_LIBRARY} evaluates to
the target name for the library named foo. In particular, it does not
guarantee that the name of the target is also foo, which the open-amp
build system currently assumes.
(It happens to be true right now, but it's an abstraction violation.)
Fix things by using the ZEPHYR_CURRENT_LIBRARY and one of its helpers.
Signed-off-by: Marti Bolivar <marti.bolivar@nordicsemi.no>
Add support for having CMAKE_BUILD_TYPE be the empty string. Setting
CMAKE_BUILD_TYPE to the empty string is valid use, so we should not
override it when it is set as such.
Also, when a multi-config generator is in use, and CMAKE_BUILD_TYPE is
set to the empty string, we will accidentally be dynamically changing
the value of CMAKE_BUILD_TYPE. This breaks the build because as the
docs say, this value is a static value and should not be dynamically
changed:
"This statically specifies what build type
https://cmake.org/cmake/help/v3.17/variable/CMAKE_BUILD_TYPE.html
This is of the same nature as
https://github.com/OpenAMP/libmetal/pull/108
Signed-off-by: Sebastian Bøe <sebastian.boe@nordicsemi.no>
Fix a compilation error due to a non-existent header file. This
was due to accidentally committing part of an experimental change
to the libmetal shared memory API.
Signed-off-by: Ed Mooring <ed.mooring@linaro.org>
If no buffer is available _rpmsg_virtio_get_buffer_size returns 0.
in this case "size > avail_size" is true so we leave the function
without waiting buffer release. Patch consists in ignoring the
test if avail_size is null.
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
This provides bug fixes and enhancements to the Xilinx-specific example
code, and a couple of small fixes to the linux example code.
Signed-off-by: Ed Mooring <ed.mooring@linaro.org>
Fix errors returned by C-STAT tool:
rpmsg_retarget.c
- L136: dead cpde found
- L185: Expression `payload_size' may overflow
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
Fix errors returned by C-STAT tool:
rpmsg_virtio.c(L347) Variable `idx'and `buff_len' may be
uninitialized
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
Fix errors returned by fix IAR C-STAT tool:
elf_loader.c:
- L374 Variable `phdr' may be uninitialized
- L469 (and some others) Expression may overflow
remoteproc.c:
- L43: Overrun in strncmp with string `mem->name'. Length passed
to strncmp is 32, but string is of length <unknown size>
- L185: Variable `ret' may be uninitialized
- L498 & L696: Value assigned to variable `ret' is never used
Signed-off-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
If the memory is not normal memory, memcmp will not work.
Compare the received buffers byte by byte.
Signed-off-by: Wendy Liang <wendy.liang@xilinx.com>
