This is a meta-issue to track the process of adding some tests to cover key things which have broken in the past or which urgently need test coverage.

LibCpp2IL

• v29 full game loading tests, any Unity version 2021.3.1 - 2021.3.39 (inclusive)
  • Windows x64
• v29.1 full game loading tests, any full release of Unity 2022.1 or 2022.2, or 2022.3 prior to 2022.3.33
  • Windows x64. This actually broke recently and wasn't caught due to lack of testing, and was fixed in 3eb40ca
• v31 full game loading tests, any Unity 2021.3 version newer than 2021.3.39 (so .40 and later)
  • Windows x64
• v31.1 full game loading tests, 2022.3.34 and later, or modern unity 6 (presumed)
  • Windows x64.
    • Technically we already test this as part of V29CustomAttributeTests, as that loads a 2022.3.35 game to get attribute blobs from. Just need to double check it covers all relevant code paths.
• Mach-O parsing tests.
  • We do not need to cover all Unity versions here, there is very little version-dependent logic, but we don't test this at all right now.
  • Arm64
  • x86_64
  • Universal mach-o binary (0xCAFEBABE magic number)
• WebAssembly parsing tests.
  • Again we do not need to cover all Unity versions here but we should ensure we can load some files and be confident we won't break anything when we push a change.
• NSO files?
  • Ideally we would have tests for NSO files, but that's possibly tricky from a legal perspective.
• PE file parsing tests where those are currently lacking.
  • Most notably there's no testing for the export parser yet.
• ELF file parsing tests where those are currently lacking.
  • x86_32 binary, for relocation testing.
  • x86_64 binary, for relocation testing.
  • Pre-24.2 testing for ARM (both v7 and v8) binaries, to test the unique logic for finding code/meta reg on those binaries.
  • Export parsing tests

Cpp2IL.Core

• Tests for various KeyFunctionAddress scanners, so we can be more confident in their accuracy on a range of binaries.
  • This will likely require all the Lib tests above so we have a good range of binaries to test scanning.
  • We should manually locate and verify the functions, then check what KeyFunctionAddress scanners find against those known good addresses.
• Tests for ISIL conversion to ensure that a known sequence of instructions results in a fixed set of ISIL.
  • x86_64
  • NewArmV8 (Disarm)
  • Other architectures as they're implemented...
• Tests for v29 attribute parsing where coverage is lacking
  • Parsing attributes with fields set (we only test parsing those with properties right now)
  • Parsing attributes that set a field or property in the base type of the attribute.
    • This will likely require a custom setup to be defined in a test game.
• Tests for MethodAnalysisContext explicit interface overrides list.
• Tests for IsPrimitive and Type on TypeAnalysisContext
  • These were discovered recently to have been broken for quite some time, see e4ffa0b and 6f9cec5