Some clang interpreter tests are failing on Windows if vcvarsall isn't run

[nico]

See http://45.33.8.238/win/89424/step_6.txt

-- Testing: 20527 tests, 48 workers --
...
C:\src\llvm-project\out\gn\obj\clang\unittests\Interpreter\.\ClangReplInterpreterTests.exe --gtest_filter=InterpreterTest.Value
--
JIT session error: Symbols not found: [ ??3@YAXPEAX_K@Z ]

This used to work back in March (and for years before). I'm currently bisecting; current regression range is 675e7bd...434ee06

[llvmbot]

@llvm/issue-subscribers-orcjit

Author: Nico Weber (nico)

See http://45.33.8.238/win/89424/step_6.txt

-- Testing: 20527 tests, 48 workers --
...
C:\src\llvm-project\out\gn\obj\clang\unittests\Interpreter\.\ClangReplInterpreterTests.exe --gtest_filter=InterpreterTest.Value
--
JIT session error: Symbols not found: [ ??3@<!-- -->YAXPEAX_K@<!-- -->Z ]

This used to work back in March (and for years before). I'm currently bisecting; current regression range is 675e7bd...434ee06

[nico]

new range: 675e7bd...6552af1

[nico]

new range: 675e7bd...4351787

[nico]

new range: 446f66d...4351787

[nico]

new range: 7630775...4351787

[nico]

$ demumble '??3@YAXPEAX_K@Z'
void __cdecl operator delete(void *, unsigned __int64)

Hm, probably 130e93c . Should we turn this off in clang-interp?

@tbaederr, looks like you touched clang-interp some. Do you have an opinion here?

[nico]

new range: d53c6cd...130e93c

[nico]

Yup, bisect finished, and it's indeed due to 130e93c

[tbaederr]

That's not the kind of interpreter I've been touching

[nico]

😅 Oh, apologies!

[wangpc-pp]

What does the vcvarsall isn't run mean here? (Not familiar with Windows...)