Review Catch2 devel through 08092139210881f01a34e60c5f2fd6afa0ea4024, the commit of 3.16.1, and apply the eleven changes since the last reviewed revision in upstream order: d7683484 better message when a reporter cannot be initialized ea5020f0 custom argument validation in every first-party reporter aa01559e throwing-reporter test also throws in assertionEnded 45cd3987 noexcept on handleUnexpectedInflightException 6adbd539 MSVC optimizations enabled again after the benchmark headers e87f2882 no quadratic scaling for NTTP test registration 73d546cd REQUIRE expands its arguments fewer times under GCC and Clang b710039a decomposer operators as plain functions f899467b type-erased streaming of reconstructed expressions 8f0e70ca retired macos-14 runner dropped 08092139 version 3.16.1 and its release notes All of them applied to the sources without a conflict. The amalgamated distribution under extras/ conflicted because it is generated; it is generated again from the sources of this repository, which also brings it up to date with fork changes it had not picked up: the list-based stream pool, the Clang diagnostic boundary and the cassert include. No file under src/Progmasoft changes. The fork-owned texts that name the engine version say 3.16.1, and docs/upstream-status.md records the review. Upstream files keep their names, licenses and compiler options. Built with Clang 22 on Windows and tested: 148 of 149 tests pass. The one that fails, Reporters::CrashInJunitReporter, fails the same way on devel without this change with that compiler.
Similar to the change that removed `std::set` from the test registration path, this saves an allocation per test to run, and is faster to boot. Technically, the performance of selecting tests for multiple OR-joined filters, each of which selects many (most) tests is a bit slower, but that is not very common use case.
This is _technically_ a performance improvement, but in actuality it is a tiny cleanup so it stops bothering me.
We used to store a full vector of "views" of the registered test infos, so that it can be retrieved externally. However, the only user of that interface left was `Session`, specifically `Session::applyFilenameAsTags`. By moving the adding of the filename tags into the `ITestCaseRegistry` interface, we avoid storing the vector and make the initial test registration faster. The new code is ugly and shows a design/layering issue, but no worse than the previous approach did.
The old internals reached into the global hub to stash the allocation(s) for enum value -> string value there, then kept around a potentially invalid (in case the hub was cleaned up) reference into it, going through whole bunch of virtual dispatch in the process. The new internals just store the data in a static variable inside the `StringMaker` specialization. This avoids potential lifetime issues, avoids all virtual dispatch and (almost) reduces the include bloat in the main header path. The reason for (almost) there is that for full include correctness, `EnumInfo` needs `<utility>` include for `std::pair`. However, this brings in things like `std::relops`, because the std headers in C++ are dumb. As this was not included before, and instead we relied on `std::pair` existing in an internal stdlib header that we transitively included, the full include size ends up bigger than before.
This tells Catch2 to create an empty file at specified path before the tests start, and delete it after the tests finish. This allows callers to catch cases where the test binary silently exits before finishing (e.g. via call to `exit(0)` inside the code under test), by looking whether the file still exists. Closes #3020