[cfe-dev] XFAIL and XTARGET

[Daniel Dunbar]

On Wed, Oct 21, 2009 at 7:26 PM, John Thompson
<john.thompson.jtsoftware at gmail.com> wrote:
> How are XFAIL and XTARGET supposed to work in a test?

For Clang they behave differently than for LLVM, this is a historical
wart. It's on my list to move Clang to the same model that LLVM uses.

> It would be nice if these were documented on the Testing Infrustructure
> page.

Yes, yes it would.

> In looking at the TestRunner.py file, it seems that you should be able to
> put something this this in a test to tell it that the test is expected to
> fail on Windows:
>
> // XFAIL: i686-pc-win32,x86_64-pc-win32
>
> But it seems that regardless of what I put after XFAIL, the test comes out
> as an expected failure if the test fails, or an unexpected pass if the test
> passes.  I've been bouncing between Linux and Windows trying different
> things.  It seems that only "// XFAIL" is currently used.

Right, Clang just looks for XFAIL and doesn't have any support for
failing on a particular target, or failing on all except a particular
target.

> I don't know Python enough to really understand what should be happening.
>
> Basically, the Driver/hello.c test doesn't pass on Windows.  I though if I
> put MinGW gcc in the PATH it might pass, but I get the following errors from
> just running the compile command line:
>
> C:\Tools\llvm\tools\clang\test>clang -ccc-echo -v -o tmp.o Driver/hello.c
> clang version 1.1 (trunk 84674)
> Target: i686-pc-win32
> Thread model: posix
>  "C:/Tools/llvm/bin/Debug/clang-cc.exe" -triple i686-pc-win32 -S
> -disable-free -main-file-name hello.c --relocation-model static --disable-f
> p-elim --unwind-tables=0 --fmath-errno=1 -v -fexceptions=0
> -fdiagnostics-show-option -o C:/DOCUME~1/fcadmin/LOCALS~1/Temp/cc-000000.s
> -x c D
> river/hello.c
> clang-cc version 1.1 based upon llvm 2.7svn hosted on i686-pc-win32
> ignoring nonexistent directory "c:\Program Files\Microsoft Visual Studio
> 9.0\VC\PlatformSDK\Include"
> ignoring nonexistent directory "/usr/local/include"
> ignoring nonexistent directory "/System/Library/Frameworks"
> ignoring nonexistent directory "/Library/Frameworks"
> #include "..." search starts here:
> #include <...> search starts here:
>  C:/Tools/llvm/bin/lib/clang/1.1/include
>  c:\Program Files\Microsoft Visual Studio 9.0\VC\include
>  /usr/include
> End of search list.
>  "c:/mingw/bin/gcc.exe" -v -c -o
> C:/DOCUME~1/fcadmin/LOCALS~1/Temp/cc-000001.o -x assembler
> C:/DOCUME~1/fcadmin/LOCALS~1/Temp/cc-000000.s
> Using built-in specs.
> Target: mingw32
> Configured with: ../gcc-4.4.0/configure
> --enable-languages=c,ada,c++,fortran,java,objc,obj-c++
> --disable-sjlj-exceptions --enable-shared --e
> nable-libgcj --enable-libgomp --with-dwarf2 --disable-win32-registry
> --enable-libstdcxx-debug --enable-version-specific-runtime-libs --prefi
> x=/mingw --with-gmp=/mingw/src/gmp/root --with-mpfr=/mingw/src/mpfr/root
> --build=mingw32
> Thread model: win32
> gcc version 4.4.0 (GCC)
> COLLECT_GCC_OPTIONS='-v' '-c' '-o'
> 'C:/DOCUME~1/fcadmin/LOCALS~1/Temp/cc-000001.o' '-mtune=i386'
>  c:/mingw/bin/../lib/gcc/mingw32/4.4.0/../../../../mingw32/bin/as.exe -o
> C:/DOCUME~1/fcadmin/LOCALS~1/Temp/cc-000001.o C:/DOCUME~1/fcadmin/L
> OCALS~1/Temp/cc-000000.s
> C:/DOCUME~1/fcadmin/LOCALS~1/Temp/cc-000000.s: Assembler messages:
> C:/DOCUME~1/fcadmin/LOCALS~1/Temp/cc-000000.s:4: Error: no such instruction:
> `align 16'
> C:/DOCUME~1/fcadmin/LOCALS~1/Temp/cc-000000.s:20: Error: no such
> instruction: `align 16'
> C:/DOCUME~1/fcadmin/LOCALS~1/Temp/cc-000000.s:22: Error: no such
> instruction: `db "I'm a little driver, short and stout.\000"'
> clang: error: assembler command failed with exit code 1 (use -v to see
> invocation)

Dunno, what's going on here.

By the way, I would recommend filing some these things as bugzillas
instead of just emailing them to cfe-dev, its makes them easier to
track over time and makes sure they don't get lost.

 - Daniel