Skip compilers that can never build Catch2 in the nightly library build

[mattgodbolt]

The nightly/catch2 job fails every day upstream Catch2 has new commits (e.g. runs on 21, 23, 24, 25 Aug: 217 packages built OK, 0 skipped, and 57 failed build) and "passes" with 0 built, 270 skipped on days it doesn't. The first errors in the log are always gcc 4.8.5 failing on operator"" _sr, which is what the daily audit has been reporting as a catch2/compiler compatibility problem. It isn't one: Catch2 v3 requires C++14 (target_compile_features(Catch2 PUBLIC cxx_std_14)), and _sr is an ordinary constexpr UDL that hasn't changed upstream since 2025.

This adds skip_compilers (precedent: spdlog) for the compilers that cannot build Catch2 for reasons intrinsic to Catch2, i.e. the ones that fail on every one of their configs and always will:

compiler	why
g485, g510, g550	no usable C++14: operator"" _sr, std::remove_reference_t, operator>> errors
avrg1610, m68kg1610, shg1610	freestanding: fatal error: ctime: No such file or directory, type_traits likewise

That's 9 of the 57 failures. I deliberately left the rest alone because they are not about Catch2 and hiding them per library would be the wrong place:

• 33 failures are the x86 + libc++ config on every clang 13 to 22, trunk, the assertion builds, armv7/armv8 clang: fatal error: '__config_site' file not found. The x86_64 libc++ and both libstdc++ configs of the same compilers build fine, so skip_compilers can't express it (disable_clang_32bit/disable_clang_libcpp are per compiler, not per combination). fmt fails the same way. A generic "does this compiler have a 32-bit libc++" probe in library_builder.py would fix it for every library.
• g63, g72 (gcc_ld-fix builds): cmake's compiler test can't find crt1.o/crti.o. Environment.
• arm1031_10, armug1520, armug1610: cmake compiler test fails at link. Cross toolchains without a sysroot set up for cmake.
• aocc520: 32-bit configs only, lib32/libstdc++.so is incompatible with elf64-x86-64.
• edg-6_8-default-15: cmake rejects target_compile_features; edg-experimental-reflection: eccp: cannot open object file.

So the job will still go red on rebuild days after this; the noise it removes is the misleading part at the top of the log. Happy to follow up on the libc++/x86 probe separately if that's wanted, since it accounts for most of what's left.

Checked with bin/ce_install --enable nightly list libraries/c++/nightly/catch2 that the entry still loads; I have no local conan setup to run the build itself.

🤖 Generated with Claude Code

[Copilot]

Copilot wasn't able to review any files in this pull request.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.