[Dexter] Replace clang with clang++ in various cross project tests

[SLTozer]

This patch attempts to fixup a set of tests in the cross-project-tests directory by replacing invocations of clang with clang++ for a set of c++ files.

As a small additional change, this patch removes -lstdc++ from a test that did not appear to require it.

[adrian-prantl]

This change looks good by itself, but we may need to additionally sneak in -isysroot if any system headers are needed by the tests.

[SLTozer]

This change looks good by itself, but we may need to additionally sneak in -isysroot if any system headers are needed by the tests.

That sounds like a good candidate for some kind of system-dependent substitution; though I don't think any of the cross-project-tests right now (including the system-darwin-specific ones) are using -isysroot, so it might not be a necessary addition.

I also see that some of the tests that were XFAILed before due to greendragon test failures don't make use of -lstdc++ and are using the correct versions of clang, so this patch likely won't fix them - will continue to investigate those cases, and add a new PR for any further fixes that fall out.

[rastogishubham]

I tried to add -isysroot to the clang invocation by modifying lit.local.cfg, that takes care of error I saw in the clang invocation, but the tests don't pass.

FAIL: cross-project-tests :: debuginfo-tests/dexter-tests/ctor.cpp (2 of 2)
******************** TEST 'cross-project-tests :: debuginfo-tests/dexter-tests/ctor.cpp' FAILED ********************
Script:
--
: 'RUN: at line 4';   /Users/shubham/Development/llvm-project/build_ninja/bin/clang -isysroot /Users/shubham/apple-internal/Xcode-Rainbow/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX14.0.sdk  -std=gnu++11 -O0 -glldb /Users/shubham/Development/llvm-project/cross-project-tests/debuginfo-tests/dexter-tests/ctor.cpp -o /Users/shubham/Development/llvm-project/build_ninja/projects/cross-project-tests/debuginfo-tests/dexter-tests/Output/ctor.cpp.tmp
: 'RUN: at line 5';   "/opt/homebrew/opt/python@3.11/bin/python3.11" "/Users/shubham/Development/llvm-project/cross-project-tests/debuginfo-tests/dexter/dexter.py" test --lldb-executable "/Users/shubham/Development/llvm-project/build_ninja/bin/lldb" --fail-lt 1.0 -w      --binary /Users/shubham/Development/llvm-project/build_ninja/projects/cross-project-tests/debuginfo-tests/dexter-tests/Output/ctor.cpp.tmp --debugger 'lldb' -- /Users/shubham/Development/llvm-project/cross-project-tests/debuginfo-tests/dexter-tests/ctor.cpp
--
Exit Code: 2

Command Output (stdout):
--
ctor.cpp: (0.0000)


--

Weirdly enough, when I run the same command line locally, it seems to work without an issue:

shubham@Mac-3004D6-2 cross-project-tests % /Users/shubham/Development/llvm-project/build_ninja/bin/clang -isysroot /Users/shubham/apple-internal/Xcode-Rainbow/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX14.0.sdk  -std=gnu++11 -O0 -glldb /Users/shubham/Development/llvm-project/cross-project-tests/debuginfo-tests/dexter-tests/ctor.cpp -o /Users/shubham/Development/llvm-project/build_ninja/projects/cross-project-tests/debuginfo-tests/dexter-tests/Output/ctor.cpp.tmp
shubham@Mac-3004D6-2 cross-project-tests % "/opt/homebrew/opt/python@3.11/bin/python3.11" "/Users/shubham/Development/llvm-project/cross-project-tests/debuginfo-tests/dexter/dexter.py" test --lldb-executable "/Users/shubham/Development/llvm-project/build_ninja/bin/lldb" --fail-lt 1.0 -w      --binary /Users/shubham/Development/llvm-project/build_ninja/projects/cross-project-tests/debuginfo-tests/dexter-tests/Output/ctor.cpp.tmp --debugger 'lldb' -- /Users/shubham/Development/llvm-project/cross-project-tests/debuginfo-tests/dexter-tests/ctor.cpp
ctor.cpp: (1.0000)

shubham@Mac-3004D6-2 cross-project-tests % echo $?
0

So I am not sure of what is going on

[rastogishubham]

@SLTozer We managed to track down the real issue and the change is in #66299

[MaskRay]

Remove -lstdc++

[rastogishubham]

Since the issue had nothing to do with clang++, is this change needed? @SLTozer @adrian-prantl

[SLTozer]

Since the issue had nothing to do with clang++, is this change needed? @SLTozer @adrian-prantl

It's not needed to make the tests pass, but we should probably be invoking the right driver for tests where reasonable, so I think it's still worth landing now that it's done.

[TiborGY]

@SLTozer Was there a reason why this PR never got merged?

[SLTozer]

Fell off my radar - it's not needed to fix any actual problems, it's just slightly better practice. I can't monitor the buildbots right now, but I'll merge it when I'm next able.