Set `_CATCH_DISCOVER_TESTS_SCRIPT` helper variable globally. Otherwise in a
scoped call (like `add_subdirectory()`) the variable gets lost. This lost
variable results in a post build error with not much information to lead to the
root of the problem.
This enables the usage of the helper script with the following example structure
- CMakeLists.txt (project root with `add_subdirectory(external/catch2)`
- external/catch2
- CMakeLists.txt (contents listed below)
- contrib/Catch.cmake
- contrib/CatchAddTests.cmake
- catch2/catch.hpp
- tests
- CMakeLists.txt (add tests with `catch_discover_tests(${PROJECT_NAME})`)
contents of project specific helper `external/catch2/CMakeLists.txt`
```cmake
cmake_minimum_required (VERSION 3.1...${CMAKE_VERSION})
project(Catch2 LANGUAGES CXX VERSION 2.13.3)
add_library(Catch2 INTERFACE)
target_include_directories(Catch2
INTERFACE
$<BUILD_INTERFACE:${CMAKE_CURRENT_LIST_DIR}>
)
# provide a namespaced alias for clients to 'link' against if catch is included as a sub-project
add_library(Catch2::Catch2 ALIAS Catch2)
include(contrib/Catch.cmake)
```
* Fixed problem with older compilers by explicitly installing the
right compiler version.
* Split apart simple builds that can be well described by a build
matrix, and builds that need special CMake defines or similar
special snowflake handling.
* Ported some extras + examples builds from TravisCI to GitHub
Actions.
This commit fixes issue that happens if the project above us uses the same variable name, thus confusing our script which see the variable scoped from the project including Catch2, rather than ours
See #2202
This test tends to be brittle on Mac CI machines, which are
heavily loaded and bursty. Since the tests are only run as part
of the "extra tests" test set, this increase should not have
a significant impact on the total duration of CI runs.
A surrogate TU is TU that includes 1 specific header, and does
nothing else. This is useful to verify that the header is
self-sufficient and does not require other headers to be included
for compilation to succeed.
Closes#2106Closes#2166 (this is a better solution)
Also generalized the implementations to write to the provided
output stream, which will be required for the follow up changes,
where the listings should happen to the location user asked for
by specifying the `-o` flag.
Previously, every base derived from the IStreamingReporter had
its own `IConfig const* m_config` member, so this just centralizes
the handling thereof.
Part of #2061
* Update cmake-integration.md
CMake related, mainly more modern and provide an executable to be correct
Co-authored-by: Martin Hořeňovský <martin.horenovsky@gmail.com>
This includes
* Testing both positive and negative path through the matchers
* Testing them with types whose `begin` and `end` member functions
require ADL
* Testing them with types that return different types from `begin`
and `end`
The old name was a legacy of v2 era, where all headers were stitched
into one. With v3 using separate headers, it is better when they have
proper name.
