Fix Catch2 XML format compatibility

[cschreib]

Originally contributed by @CrustyAuklet in #169. Rebased on the latest main branch and refactoring to solve the remaining issues.

Also fixes #179.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 94.17%. Comparing base (ecfa3d6) to head (c8bc7e0).
Report is 24 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #183      +/-   ##
==========================================
+ Coverage   94.03%   94.17%   +0.13%     
==========================================
  Files          29       30       +1     
  Lines        1709     1750      +41     
==========================================
+ Hits         1607     1648      +41     
  Misses        102      102              

Flag	Coverage Δ
	94.17% <100.00%> (+0.13%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
include/snitch/snitch_registry.hpp	86.36% <ø> (ø)
src/snitch_capture.cpp	92.00% <ø> (ø)
src/snitch_registry.cpp	95.58% <100.00%> (+0.39%)	⬆️
src/snitch_reporter_catch2_xml.cpp	95.45% <ø> (-0.44%)	⬇️
src/snitch_reporter_console.cpp	97.14% <ø> (ø)
src/snitch_reporter_teamcity.cpp	100.00% <100.00%> (ø)
src/snitch_section.cpp	88.46% <100.00%> (+2.74%)	⬆️
src/snitch_test_data.cpp	95.83% <100.00%> (+0.18%)	⬆️
src/snitch_time.cpp	100.00% <100.00%> (ø)

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ecfa3d6...c8bc7e0. Read the comment docs.

[cschreib]

Trying to add more tests to catch the lines currently missing from the coverage; I noticed there's an edge case to the current implementation:

struct destructor_asserter {
    ~destructor_asserter() {
        CHECK(false);
    }
};

TEST_CASE("test") {
    SECTION("section 1") {
        destructor_asserter a;
        SECTION("section 2") {
            throw std::runtime_error("oops");
        }
    }
}

This is actually a pre-existing issue introduced in #178.

This currently causes us to miss the section_ended event for "section 2", because the "held" section state (created when an uncaught exception is detected) is currently cleared unconditionally whenever we encounter a new CHECK() (or SECTION(), or CAPTURE(), or INFO()). I think there's a way to address this, but I wonder if I will have to rethink the exception handling logic for sections... Will pick it again a bit later.

[cschreib]

Solved by reworking the "held state clearing" logic. This is no longer done unconditionally on the first CHECK*()/CAPTURE()/INFO()/SECTION() encountered (which was the implementation from #178), but instead we detect when the stack is being unwinded because of an exception, and skip the clearing if so. Otherwise, we close any section that was left open in the "held state", and then clear the held state. This solves the unbalanced section events.

We also expose this logic in a new public function snitch::notify_exception_handled();. This lets users trigger the clearing manually when they think the time is right (e.g., in their custom catch handler). This solves #179 in passing.

TEST_CASE("foo") {
    try {
        // Your test code that can throw...
    } catch (...) {
        // Your custom handling logic...
        
        // This is now necessary:
        snitch::notify_exception_handled();
    }

    // Now, all following code works fine.
}

If that call is forgotten, most things will still work fine: only a following unhandled exception would have confusing contextual information (i.e., wrong capture/info/section data reported). More detailed explanations are in #179.

[cschreib]

I'm happy to merge this. @CrustyAuklet if you are able to test this before merge, it would be awesome! Otherwise I'll just merge it sometime this weekend, and we can fix any issues in a follow up PR. Thanks for your help regardless, and sorry this took so long to merge...