[Driver] move FreeBSD header search path management to the driver

[mhjacobson]

Summary

This matches OpenBSD, and it supports Swift's use of clang for its C interop functionality. Recent changes to Swift use AddClangSystemIncludeArgs() to inspect the cc1 args; this doesn't work for platforms where cc1 adds standard include paths implicitly.

Also, clean up InitHeaderSearch, making it clearer which targets manage header search paths in the driver.

Add test.

[mhjacobson]

Added test.

[mhjacobson]

Run clang-format.

[mhjacobson]

Clarify comment.

[MaskRay]

clang/lib/Driver/ToolChains/FreeBSD.cpp

442

I think Fuchsia way of checking if (!D.SysRoot.empty()) { before adding /usr/include is probably better. Is it /include or /usr/include ?

clang/lib/Lex/InitHeaderSearch.cpp

103

Drop ShouldAddDefaultIncludePaths - . It is not recommended for new code.

104

multiple spaces => one space

clang/test/Driver/freebsd.c

217

Is this sufficient? For Linux linux-cross.cpp checks a lot of directories to facilitate code refactoring.

[mhjacobson]

clang/lib/Driver/ToolChains/FreeBSD.cpp

442

It's /usr/include on FreeBSD.

I'm confused by the Fuchsia code—does it not add /include if the sysroot is empty? Appears not...

$ clang-15 -### -c -target arm64-fuchsia -xc /dev/null         
clang version 15.0.3
Target: arm64-unknown-fuchsia
Thread model: posix
InstalledDir: /opt/local/libexec/llvm-15/bin
 (in-process)
 "/opt/local/libexec/llvm-15/bin/clang" "-cc1" "-triple" "arm64-unknown-fuchsia" "-emit-obj" "-mrelax-all" "--mrelax-relocations" "-disable-free" "-clear-ast-before-backend" "-disable-llvm-verifier" "-discard-value-names" "-main-file-name" "null" "-mrelocation-model" "pic" "-pic-level" "2" "-pic-is-pie" "-mframe-pointer=non-leaf" "-ffp-contract=on" "-fno-rounding-math" "-mconstructor-aliases" "-funwind-tables=2" "-target-cpu" "generic" "-target-feature" "+neon" "-target-feature" "+v8a" "-target-abi" "aapcs" "-fallow-half-arguments-and-returns" "-mllvm" "-treat-scalable-fixed-error-as-warning" "-debugger-tuning=gdb" "-target-linker-version" "556.6" "-fcoverage-compilation-dir=/Users/matt" "-resource-dir" "/opt/local/libexec/llvm-15/lib/clang/15.0.3" "-I/usr/local/include" "-internal-isystem" "/opt/local/libexec/llvm-15/lib/clang/15.0.3/include" "-fdebug-compilation-dir=/Users/matt" "-ferror-limit" "19" "-fsanitize=shadow-call-stack" "-stack-protector" "2" "-fno-signed-char" "-fgnuc-version=4.2.1" "-fcolor-diagnostics" "-target-feature" "+outline-atomics" "-faddrsig" "-D__GCC_HAVE_DWARF2_CFI_ASM=1" "-o" "null.o" "-x" "c" "/dev/null"

$ clang-15 -### -c -target arm64-fuchsia --sysroot /foo -xc /dev/null
clang version 15.0.3
Target: arm64-unknown-fuchsia
Thread model: posix
InstalledDir: /opt/local/libexec/llvm-15/bin
 (in-process)
 "/opt/local/libexec/llvm-15/bin/clang" "-cc1" "-triple" "arm64-unknown-fuchsia" "-emit-obj" "-mrelax-all" "--mrelax-relocations" "-disable-free" "-clear-ast-before-backend" "-disable-llvm-verifier" "-discard-value-names" "-main-file-name" "null" "-mrelocation-model" "pic" "-pic-level" "2" "-pic-is-pie" "-mframe-pointer=non-leaf" "-ffp-contract=on" "-fno-rounding-math" "-mconstructor-aliases" "-funwind-tables=2" "-target-cpu" "generic" "-target-feature" "+neon" "-target-feature" "+v8a" "-target-abi" "aapcs" "-fallow-half-arguments-and-returns" "-mllvm" "-treat-scalable-fixed-error-as-warning" "-debugger-tuning=gdb" "-target-linker-version" "556.6" "-fcoverage-compilation-dir=/Users/matt" "-resource-dir" "/opt/local/libexec/llvm-15/lib/clang/15.0.3" "-isysroot" "/foo" "-internal-isystem" "/opt/local/libexec/llvm-15/lib/clang/15.0.3/include" "-internal-externc-isystem" "/foo/include" "-fdebug-compilation-dir=/Users/matt" "-ferror-limit" "19" "-fsanitize=shadow-call-stack" "-stack-protector" "2" "-fno-signed-char" "-fgnuc-version=4.2.1" "-fcolor-diagnostics" "-target-feature" "+outline-atomics" "-faddrsig" "-D__GCC_HAVE_DWARF2_CFI_ASM=1" "-o" "null.o" "-x" "c" "/dev/null"

I don't know anything about Fuchsia, but isn't that wrong? Certainly for FreeBSD we'd want to add an unprefixed /usr/include normally...

[mhjacobson]

clang/test/Driver/freebsd.c

217

On FreeBSD, the only directories included by default are

• /usr/include/c++/v1 (in C++ mode)
• <resource dir>/include
• /usr/include

I suppose I can add checks for the first two.

[mhjacobson]

Test for more paths; test C++.

[mhjacobson]

Fix a couple warnings.

[MaskRay]

clang/lib/Lex/InitHeaderSearch.cpp

104

not done. /// => ///

234

drop braces around simple single statements

391

move the variable immediately before switch os and move the comment before the variable? Avoid unneeded blank lines

clang/test/Driver/freebsd.cpp

48 ↗

(On Diff #476042)

Place "-internal-isystem on the same line to ensure there are no extra includes in between.

Alternatively, use the -NEXT: {{^}} xxx pattern I use in linux-cross.cpp

[mhjacobson]

Style fixes.

Ensure include paths are sequential.

[mhjacobson]

Make the tests play nice with Windows path separators.

[MaskRay]

This matches OpenBSD, and it supports Swift's use of clang for its C interop functionality. Recent changes to Swift use AddClangSystemIncludeArgs() to inspect the cc1 args; this doesn't work for platforms where cc1 adds standard include paths implicitly.

Use a commit link if known.

Also, clean up InitHeaderSearch, making it clearer which targets manage header search paths in the driver.

Add test.

"Add test" is unneeded in the commit message:)