| Age | Commit message (Collapse) | Author |
|---|
|
The pull request #88 was reverted as it broke travis build.
In fact it only reveled a bug that was introduced in #80 (1b071fadd).
The previous commit 0c0cc95 fixes the problem (cmake variable shadow).
Revert the revert (d17b931a3dcc70cf9e) to restore the pull request #88.
|
|
The pull request #88 was reverted as it broke travis build.
ctest uses a cache variable MEMCHECK_SUPPRESSION_FILE.
It sets it during the `include(ctest)` evaluation.
Cache variable are set on creation, they can not be modify after by a
CMakeList.mk (except with set's FORCE option).
ef5736eb5 sets MEMCHECK_SUPPRESSION_FILE as a normal variable.
This does not set the cache variable as expected but rather create
a second local variable (not part of the cache) with the same name.
If cmake was run once (as during travis build since patch 4e41ff7,
until its revert d17b931), the local variable has no effect
thus the suppression file was never saved in the cache.
If cmake was run twice, during the second time, as the cache version
is not recreated (cache variables are persistent),
the call to set MEMCHECK_SUPPRESSION_FILE changes the cache version as
it was first expected.
As a result the build worked if cmake was invoked twice but failed
if only called once.
For more information, see the *** Variable types in CMake *** section at:
http://www.cmake.org/cmake/help/v2.8.11/cmake.html#command%3aset
This patch fixes all that by not using MEMCHECK_SUPPRESSION_FILE.
but rather setting directly the option in valgrind.
It is a clean fix as the MEMCHECK_SUPPRESSION_FILE is an option
for the user to add a valgrind suppression file, it is not
intended to be used by the project direcly.
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
This reverts commit 339779ac82acb755f6e992bf7ee54e3b791a9ca6, reversing
changes made to 1b071faddaf8c5e835ea38d6c8d2d5db88011251.
|
|
Debug build used to be installed as it was required
for make test.
As make test is now stand alone, do not install debug builds.
This has several advantages:
- slightly faster builds.
- make test without install is actually tested by travis.
Debug binaries used to overwrite release ones that were installed
presciently. As it is no longer the case, and to avoid tests to use
installed binaries instead of the local ones, test the debug build
*before* the release one.
As the debug build is now build (first) in it's own folder,
debug and release build can not interfere.
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
Long option are more readable.
|
|
With a classic cmake install prefix '/usr',
bash completions were installed in
/usr/etc/bash_completion.d/
which even before bash-completion 1.9 was incorrect.
Since v1.9 all completions are to be installed in
/usr/share/bash-completion/completions
See:
http://anonscm.debian.org/cgit/bash-completion/bash-completion.git/tree/CHANGES#n324
Change the completions install path accordingly
|
|
Gcov was run twice, first by the target ExperimentalCoverage
and then by coveralls.
As coveralls give a better coverage report and cdash is not used,
remove ExperimentalCoverage.
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
Some unit test were checked using valgrind,
but if an error was detected it was only logged:
the build was not failed.
Fail to build all *MemCheck targets if
a leak or an error is reported by valgrind.
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
Mainly add a some documentation about the custom cmake
set_test_env function.
|
|
Ctest related files and definitions were at the
source root although we try to keep it simple.
Move all ctest related files and definition in
a ctest directory.
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
The root CMakeList was including every test directories
directly, although it should be kept simple.
Move the inclusions to a dedicated CMakeList in test.
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
coveralls.io is a service that uses information from gcov (generated during
"make test") to provide a coverage report online:
https://coveralls.io/r/01org/parameter-framework
Some files must be ignored, specifically: generated files (python bindings and
files generated by CMake), tests (because they would inflate the coverage rate)
and samples (for now, only the skeleton subsystem).
A "badge" is also added to the readme showing the coverage rate.
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
Backsynchronizer is dynamically allocated for simulation purpose.
Nevertheless, this is useless as this can be easily achieved thanks to a
typedef.
This patch simplify the backsynchronization use.
Signed-off-by: Sébastien Gonzalve <oznog@users.noreply.github.com>
|
|
On windows platform, the definition of a function pointer to a member
of a given class requires visibility on the header of that class,
otherwize memory corruption may occur at deletion time.
Change-Id: Ifa4943ef0f3e02542c6641959cf67384103bfc69
Signed-off-by: Patrick Benavoli <patrick.benavoli@intel.com>
|
|
Remove misleading hungarian notation of variable name in
CRuleParser::iterate() that was making the code hard to read/maintain.
Change-Id: I7a034d7999f419263551c3ed9a7c26cda38e3818
Signed-off-by: Sebastien Gonzalve <sebastien.gonzalve@intel.com>
|
|
This patch removes the type mismatch warnings revealed by Windows 64
compiler. Wherever necessary, used size_t type for size related data.
Change-Id: Ie045ce95940cd83fe8d681168ac9526fc6028d43
Signed-off-by: Patrick Benavoli <patrick.benavoli@intel.com>
|
|
Made _bValidateSchemasOnStart attribute initialized as part of the
constructor's initializer list.
Change-Id: I8f345b5c17f33e012b69005e01dc31a270e41778
Signed-off-by: Patrick Benavoli <patrick.benavoli@intel.com>
|
|
Those functions are used in one place and let the caller transform a
std::string to a char * which makes the use of alloca necessary.
Moreover, the readString cannot know if the pointer given in input
points to a buffer with enough space, making it dangerous.
This patch reworks the API so that write/readString() get std::string
and lets the implementation play with char *. This also allows to get
rid of alloca() use which is quite dangerous.
Change-Id: I973e4987278289bb914c4e194b86e924a1c88a6d
Signed-off-by: Sebastien Gonzalve <sebastien.gonzalve@intel.com>
|
|
Some xsd files did not have linux style line endings.
Change-Id: Id2ec46bf740679e5736bd7283377c610f7c320f9
Signed-off-by: Sebastien Gonzalve <sebastien.gonzalve@intel.com>
|
|
Spurious execution mode removed from files:
- parameter/Element.cpp
- utility/convert.hpp
Change-Id: I5d47d505e8c3d9f55ad975c34b51d7caa4905ab9
Signed-off-by: Sebastien Gonzalve <sebastien.gonzalve@intel.com>
|
|
CElementBuilder constructor had a ";" right after its defintion.
This patch removes it.
Change-Id: I60551bf9610b1760181d9b206a98f566cef78f0b
Signed-off-by: Sebastien Gonzalve <sebastien.gonzalve@intel.com>
|
|
File removed from remote-process folder
Change-Id: Iec37d2581fd424e7cfd891224813e57a56caa2a6
Signed-off-by: Sebastien Gonzalve <sebastien.gonzalve@intel.com>
|
|
Added #pragma once on top of the file to prevent multiple inclusion.
Change-Id: I351ada48e16bf6bfe84b546da5581b1aa6bce03e
Signed-off-by: Sebastien Gonzalve <sebastien.gonzalve@intel.com>
|
|
Replace #ifdef and #define wth #pragma once
Change-Id: I60fe1e74a7481be38d9b8e44952e533d7ae80857
Signed-off-by: Sebastien Gonzalve <sebastien.gonzalve@intel.com>
|
|
Pthread was required by included directly findPthread.cmake
then testing for success.
Use the find_package command and require success with the REQUIRED
option.
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
Stopping the remote-processor in the stop command has two side effect:
- stopping the remote-processor thread will fail as it is not possible
to join a thread to itself. This leads to the leak of all the thread
object (destructor not called).
- the server will not return the exit command status as it has been
stopped during it's execution.
Remove the remote-processor stop as it will be called during it's
destruction.
Signed-off-by: Kevin Rocard <kevinx.rocard@intel.com>
|
|
When an error occurs during input parsing,
the line number was incorrectly displayed.
In python, enumerating an iterable will yield
an pair of index and a value.
This index starts at 0: next(enumerate([""])) == (0, "")
This index was displayed as the line number.
Add one to the index to get the line number.
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
Some pfw bug induce empty criterion value,
when log it is dificult to see the the empty
value.
Add quotes around values.
before: Error in criterion
After : Error in criterion ""
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
This test is failing (on travis only) for strange reasons. As we can not trust
it, let's remove it for now.
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
This patch adds functionnal test to the PFW.
The execution of these tests needs the test-subsystem to be compiled.
The tests can be launch with 'make test'.
We set the environment vars in the cmake.
Then we launch ACTCampaignEngine.py.
Temporary files like build time conf files are placed on the build directory.
Next steps :
- Test-subsystem needs to be reworked (with binding python).
Signed-off-by: Adrien M <adrienx.martin@intel.com>
|
|
Possibility to run the test without having to lunch the 'make install'
command.
Signed-off-by: Adrien M <adrienx.martin@intel.com>
|
|
This test was using the wrong array size. It leads to false positive.
Signed-off-by: Jules Clero <julesx.clero@intel.com>
|
|
This test was trying to set string parameter containing
parenthesis through remote-process without escaped them.
This patch introduces the escaped value in order to
correctly use the remote-process command through bash.
Signed-off-by: Jules Clero <julesx.clero@intel.com>
|
|
This test is comparing lower case hexadecimal to upper case
hexadecimal number. The test was failing because
0xFF is not 0xff when we compare string. As 0xff is read back
from filesystem, we now compare it to the right value.
Signed-off-by: Jules Clero <julesx.clero@intel.com>
|
|
Parameter-framework in now displaying pending configurations, we need
to take this in account in our test.
Signed-off-by: Jules Clero <julesx.clero@intel.com>
|
|
Import of parameter-framework tests.
Simplified test engine and use of python test discovery.
Signed-off-by: Herve Garanto <hervex.garanto@intel.com>
Signed-off-by: Sylvère Billout <sylverex.billout@intel.com>
Signed-off-by: Philippe Colonges <philippeX.colonges@intel.com>
Signed-off-by: Jules Clero <julesx.clero@intel.com>
|
|
When translating a numerical value to literal for an inclusive
criterion type, if the numerical value is not found into the map
of pairs, it appends anyway the "|".
This patch removes this artifacts display if the numerical value is not
part of the acceptable values of this criterion type.
Change-Id: I2ff4ee1ae6e6e3ac87c94c596bae3ae534f173c4
Signed-off-by: Francois Gaffie <francois.gaffie@intel.com>
|
|
C bindings were compiled and tested but never installed.
Add the missing install command.
Signed-off-by: Kevin Rocard <kevin.rocard@intel.com>
|
|
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
Some Android release scripts seem to need these files.
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
It didn't have a license header. Even though we received explicit authorization
to use it, it wasn't a comfortable situation.
This is an original implementation that only keeps the APIs we use in the
parameter-framework and keep them identical (except for a small exception). The
behaviour is also exactly the same.
Change-Id: I85a69c76027ee026a693d79cd19edd3b86796f9d
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
In 327b1a48e2fe539c0ca5414ce9c98b2405536a09 we changed the installation prefix
to /usr because we witnessed errors during unit tests using the python
bindings.
We initially thought that /usr/local/lib wasn't in the loader's default paths
but it turns out it is. In fact, it only is the loader's cache that was
outdated and needed to be regenerated. That's what ldconfig does.
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
Testing slightly-out-of-range values is failing (i.e. the value seems to be
wrongfully accepted by the parameter-framework) starting from high precisions,
e.g. q0.16.
The cause is:
1) When converting strings to fixed-points, the parameter-framework first
parses the input string as double. Doubles have 52 fraction bits and there for
15 decimal-digits-precision (52 * log10(2) == 15).
2) When displaying Qn.m numbers, we decided to display all representable digits
(i.e. m digits) instead of only displaying significant digits because that
would cause unpredictable behaviour. See
d299108157ee4d0eadb7683b2fa6a6635bc63d95 and
59cc1e33810c55e6fa1e3bd320e1cf29e24d23be for more information.
3) However, the precision for a Qn.m fixed-point number isn't m but
m * log10(2). This is guaranteed to be less than the precision of a double but
the tests mentioned at the beginning were written as if the precision was m,
which may be larger than the precision of a double. This is what caused the
test failures.
The tests are adapted to try and set a value of which the last significant
digit is one-off too far outside of the allowed range.
For lower-precision number, other less-far-off values could be correctly
rejected by the parameter-framework but that's only because of the underlying
implementation (using doubles as intermediate values) and it seems hardly
relevant to check.
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
We used to drop errors during "get" operation because it is unlikely to fail
and not in the scope of the fixed-point tests. However, we should catch such
error and abort immediately.
Signed-off-by: David Wagner <david.wagner@intel.com>
|
|
Using the python bindings is much faster and easier than using the network
interface.
Signed-off-by: David Wagner <david.wagner@intel.com>
|
