1
0
mirror of https://github.com/esp8266/Arduino.git synced 2025-06-04 18:03:20 +03:00
M Hightower d1d4212e8b
Add support for global build defines and options (#8504)
* Add support to specify global build defines and options

A script manages the use of a file with a unique name, like
`SketchName.ino.globals.h`, in the Sketch source directory to provide compiler
command-line options (build options) and sketch global defines. The build
option data is encapsulated in a unique "C" comment block and extracted into
the build tree during prebuild.

* Applied os.path.normpath() liberally to input arguments. Fixes windows file path issue.
Improved helpful message for adding embedded build options.

* doubleup '\'

* Added context help for build option support

* expunged sketchbook global
added workaround for aggressive caching

* inital pass at searching for and reading preferences.txt

* Correct Windows path for preferences.txt
Added portable path for preferences.txt
Expanded file timestamp granularity
Improved error message printing for Arduino IDE 2.0 RC4

* Improved portable path and various Windows paths to preferences.txt

* Add cleanup logic and identify 1st run after IDE restart

* text corrections

* Create mkbuildoptglobals.py

When global header file does not exist, this print makes it easier for user to create the header file by providing its name and documentation pointer.

* build.opt heads up to user

Compiler command line changes from build.opt are shown to user

* Updated text

* oops

* Expanded comment and made print help consistent

* Improve handling stderr/stdout with "no verbose output"
Grouped helpful info to print at the end.
Added missing return value.

* Correct timestamp on CommonHFile.h
More improvements to printing
Updated docs.

* Added command-line parser

Support hints for compiler.cache_core. For use when Arduino IDE uses
command-line options that override compiler.cache_core.

Removed overuse of ()

Improve FAQ entry

* Fix script failure under windows

Rely on argpaser for checking that all arguments are present.
Removed redundant argument check in main().

Added '--debug' option and print_dbg method.

Rethink failures on overrides. Remove well know path fallbacks,
error exit when override file is missing.

In well-known path search for preferences.txt, do not assume true.
Make failure to find an error exit event.

When Windows has two preferences.txt files and they have different
values for caching and globals.h is used, error exit. It is not
possible to know from the script which is being used.

* Use quotes on build.opt

Update comment
Include the @ within the expantion string use quotes around file name.
Update doc example to remind and use quotes.

* Update CI for build option and global support

Added "mkbuildoptglobals.extra_flags=--cache_core" to platform.loca.txt
Update "-ide-version=10802" this version number indicates aggressive caching support
Added example to test global .h support

* Add debug prints
Added --debug to CI - this needs to be removed later
Tweaks to touch...

* Give each build VM a unique build.tmp space

* Corrected style on example
temp CI changes
debug crud
Added --ci switch

* Removed CI debug crud

run_CI_locall.sh works fine locally. Hosted Multi-VM CI fails
to work with 'aggressive caching' workaround method.

Add #if defined(CORE_MOCK) to failing example.

* Try HOST_MOCK

* CI adjustments

mkbuildoptglobals.py is optimized around the Arduino IDE 1.x
behaviour. One way the CI differs from the Arduino IDE is in the
handling of core and caching core. With the Arduino IDE, each sketch
has a private copy of core and contributes to a core cache. With the
CI, there is one shared copy of core for all sketches. When global
options are used, the shared copy of core and cache are removed before
and after the build.

* Doc update
2022-05-12 17:14:17 +02:00
..
2018-02-19 17:11:48 +03:00
2019-07-03 09:49:03 +02:00
2020-09-01 23:37:14 +09:00

Testing Arduino ESP8266 Core

Testing on host

Some features of this project can be tested by compiling and running the code on the PC, rather than running it on the ESP8266. Tests and testing infrastructure for such features is located in tests/host directory of the project.

Some hardware features, such as Flash memory and HardwareSerial, can be emulated on the PC. Others, such as network, WiFi, and other hardware (SPI, I2C, timers, etc) are not yet emulated. This limits the amount of features which can be tested on the host.

Adding a test case

Tests are written in C++ using Catch framework.

See .cpp files under tests/host/core/ for a few examples how to write test cases.

When adding new test files, update TEST_CPP_FILES variable in tests/host/Makefile to compile them.

If you want to add emulation of a certain feature, add it into tests/host/common/ directory.

Running test cases

NOTE! The test-on-host environment is dependent on some submodules. Make sure to run git submodule update --init before running any test.

To run test cases, go to tests/host/ directory and run make. This will compile and run the tests.

If all tests pass, you will see "All tests passed" message and the exit code will be 0.

Additionally, test coverage info will be generated using gcov tool. You can use some tool to analyze coverage information, for example lcov:

lcov -c -d . -d ../../cores/esp8266 -o test.info
genhtml -o html test.info

This will generate an HTML report in html directory. Open html/index.html in your browser to see the report.

Note to macOS users: you will need to install GCC using Homebrew or MacPorts. Before running make, set CC, CXX, and GCOV variables to point to GCC tools you have installed. For example, when installing gcc-5 using Homebrew:

export CC=gcc-5
export CXX=g++-5
export GCOV=gcov-5

When running lcov (which you also need to install), specify gcov binary using --gcov-tool $(which $GCOV) (assuming you have already set GCOV environment variable).

Testing on device

Most features and libraries of this project can not be tested on host. Therefore testing on an ESP8266 device is required. Such tests and the test infrastructure are located in tests/device directory of this project.

Test cases

Tests are written in the form of Arduino sketches, and placed into tests/device/test_xxx directories. These tests are compiled using Arduino IDE, so test file name should match the name of the directory it is located in (e.g. test_foobar/test_foobar.ino). Tests use a very simple BSTest library, which handles test registration and provides TEST_CASE, CHECK, REQUIRE, and FAIL macros, similar to Catch.

Note: we should migrate to Catch framework with a custom runner.

Here is a simple test case written with BSTest:

#include <BSTest.h>
#include <test_config.h>

BS_ENV_DECLARE();

void setup()
{
    Serial.begin(115200);
    BS_RUN(Serial);
}


TEST_CASE("this test runs successfully", "[bs]")
{
    CHECK(1 + 1 == 2);
    REQUIRE(2 * 2 == 4);
}

BSTest is a header-only library, so necessary static data is injected into the sketch using BS_ENV_DECLARE(); macro.

BS_RUN(Serial) passes control to the test runner, which uses Serial stream to communicate with the host. If you need to do any preparation before starting tests, for example connect to an AP, do this before calling BS_RUN.

TEST_CASE macro defines a test case. First argument is human-readable test name, second contains optional set of tags (identifiers with square brackets). Currently only one tag has special meaning: [.] can be used to mark the test case as ignored. Such tests will not be skipped by the test runner (see below).

Test execution

Once BS_RUN is called, BSTest library starts by printing the menu, i.e. the list of tests defined in the sketch. For example:

>>>>>bs_test_menu_begin
>>>>>bs_test_item id=1 name="this test runs successfully" desc="[bs]"
>>>>>bs_test_menu_end

Then it waits for the test index to be sent by the host, followed by newline.

Once the line number is received, the test is executed, and feedback is printed:

>>>>>bs_test_start file="arduino-esp8266/tests/device/test_tests/test_tests.ino" line=13 name="this test runs successfully" desc="[bs]"
>>>>>bs_test_end line=0 result=1 checks=2 failed_checks=0

Or, in case the test fails:

>>>>>bs_test_start file="arduino-esp8266/tests/device/test_tests/test_tests.ino" line=19 name="another test which fails" desc="[bs][fail]"
>>>>>bs_test_check_failure line=22
>>>>>bs_test_check_failure line=24
>>>>>bs_test_end line=0 result=0 checks=4 failed_checks=2

BSTest library also contains a Python script which can "talk" to the ESP8266 board and run the tests, tests/device/libraries/BSTest/runner.py. Normally it is not necessary to use this script directly, as the top level Makefile in tests/device/ directory can call it automatically (see below).

Test configuration

Some tests need to connect to WiFi AP or to the PC running the tests. In the test code, this configuration is read from environment variables (the ones set using C getenv/setenv functions). There are two ways environment variables can be set.

  • Environment variables which apply to all or most of the tests can be defined in tests/device/test_env.cfg file. This file is not present in Git by default. Make a copy of tests/device/test_env.cfg.template and change the values to suit your environment.

  • Environment variables which apply to a specific test can be set dynamically by the setup host side helper (see section below). This is done using setenv function defined in mock_decorators.

Environment variables can also be used to pass some information from the test code to the host side helper. To do that, test code can set an environment variable using setenv C function. Then the teardown host side helper can obtain the value of that variable using request_env function defined in mock_decorators.

A SPIFFS filesystem may be generated on the host and uploade before a test by including a file called make_spiffs.py in the individual test directory.

Building and running the tests

Makefile in tests/device/ directory handles compiling, uploading, and executing test cases.

Here are some of the supported targets:

  • virtualenv: prepares Python virtual environment inside tests/device/libraries/BSTest/virtualenv/. This has to be run once on each computer where tests are to be run. This target will use pip to install several Python libraries required by the test runner (see tests/device/libraries/BSTest/requirements.txt).

  • test_xxx/test_xxx.ino: compiles, uploads, and runs the tests defined in test_xxx/test_xxx.ino sketch. Some extra options are available, these can be passed as additional arguments to make:

    • NO_BUILD=1: don't compile the test.
    • NO_UPLOAD=1: don't upload the test.
    • NO_RUN=1: don't run the test.
    • V=1: enable verbose output from compilation, upload, and test runner.

    For example, make test_newlib/test_newlib.ino V=1 will compile, upload, and run all tests defined in test_newlib/test_newlib.ino.

    For each test sketch, test results are stored in tests/device/.build/test_xxx.ino/test_result.xml. This file is an xUnit XML file, and can be read by a variety of tools, such as Jenkins.

  • test_report: Generate HTML test report from xUnit XML files produced by test runs.

  • all (or just make without a target): Run tests from all the .ino files, and generate HTML test report.

Host-side helpers

Some tests running on the device need a matching part running on the host. For example, HTTP client test might need a web server running on the host to connect to. TCP server test might need to be connected to by TCP client running on the host. To support such use cases, for each test file, an optional Python test file can be provided. This Python file defines setup and teardown functions which have to be run before and after the test is run on the device. setup and teardown decorators bind setup/teardown functions to the test with specified name:

from mock_decorators import setup, teardown, setenv, request_env

@setup('WiFiClient test')
def setup_wificlient_test(e):
    # create a TCP server
    # pass environment variable to the test
    setenv(e, 'SERVER_PORT', '10000')
    setenv(e, 'SERVER_IP', repr(server_ip))

@teardown('WiFiClient test')
def teardown_wificlient_test(e):
    # delete TCP server
    # request environment variable from the test, compare to the expected value
    read_bytes = request_env(e, 'READ_BYTES')
    assert(read_bytes == '4096')

Corresponding test code might look like this:


TEST_CASE("WiFiClient test", "[wificlient]")
{
    const char* server_ip = getenv("SERVER_IP");
    int server_port = (int) strtol(getenv("SERVER_PORT"), NULL, 0);

    WiFiClient client;
    REQUIRE(client.connect(server_ip, server_port));

    // read data from server
    // ...

    // Save the result back so that host side helper can read it
    setenv("READ_BYTES", String(read_bytes).c_str(), 1);
}