Failing tests on AArch64/RISC-V

[spholz]

These tests currently fail on AArch64 and RISC-V:

AArch64:

• AK/TestComplex: Tests from_polar_magnitude and from_polar_phase fail AK: Make sin(), cos() more precise on non-x86 #26669
• Kernel/crash: Division by zero doesn't cause a signal crash: Implement all architecture-specific tests for AArch64 #26001
• LibC/TestFenv: TODO_AARCH64s in aarch64/fenv.cpp LibC: Add basic fenv support on AArch64 #26697
• LibC/TestLibCSetjmp: Crash, {set,sigset,long}jmp aren't implemented on AArch64 LibC: Implement setjmp and longjmp for aarch64 #26000
• LibC/TestMath: Crash, internal_gamma always uses rintl, but long double math functions are unimplemented on AArch64
• LibGfx/TestDeltaE: Multiple EXPECT_APPROXIMATEs fail AK: Make atan() more precise and faster on non-x86 #26691 (probably?)
• LibPDF/TestPDF: Test postscript fails AK: Make atan() more precise and faster on non-x86 #26691
• test-js: Math.tan.js fails, hangs during builtins/Intl/NumberFormat/NumberFormat.prototype.format.js (AK: Let fmod(n*x, x) return 0 #26599)
• Crash tests cause CrashReporter popus even if they crash expectedly (maybe because the kernel always raises SIGSEGV for all hardware exceptions?) Kernel: Use an up-to-date FPUState for children created with fork() #26041
• TestEFault is currently disabled on AArch64:

  serenity/Tests/Kernel/CMakeLists.txt

  Line 74 in 93add0a

  # FIXME: Don't panic on SafeMem access faults on aarch64

  because AArch64 is missing SafeMem implementations that don't panic if the kernel fails to access userspace memory Kernel/aarch64: Actually implement safe SafeMem functions #26002

RISC-V:

• AK/TestComplex: Tests from_polar_magnitude and from_polar_phase fail AK: Make sin(), cos() more precise on non-x86 #26669
• LibC/TestMath: Crash, internal_gamma always uses rintl, but long double math functions are unimplemented on RISC-V
• LibGfx/TestDeltaE: Multiple EXPECT_APPROXIMATEs fail AK: Make atan() more precise and faster on non-x86 #26691 (probably?)
• LibPDF/TestPDF: Test postscript fails AK: Make atan() more precise and faster on non-x86 #26691
• test-js: Math.tan.js fails, hangs during builtins/Intl/NumberFormat/NumberFormat.prototype.format.js (AK: Let fmod(n*x, x) return 0 #26599)

These tests fail on RISC-V if the time wan't set by ntpquery -s (and likely also on AArch64/x86-64 if the time isn't set correctly):

• LibC/TestLibCTime: Test tzset fails Kernel/riscv64: Support the Google Goldfish real-time clock #26698
• LibTLS/TestTLSHandshake: Test test_TLS_hello_handshake fails Kernel/riscv64: Support the Google Goldfish real-time clock #26698
• test-js: Multiple Date tests fail Kernel/riscv64: Support the Google Goldfish real-time clock #26698

All tests that fail on RISC-V also fail on AArch64. These common test failures seem to all be caused by missing/inaccurate math functions.