[PATCH] D33013: Driver must return non-zero code on errors in command line

Richard Smith via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Tue May 9 14:44:38 PDT 2017

rsmith added a comment.

Thank you, some of these test typos are ... alarming. =)

A couple of the test updates don't look quite right, but this mostly looks great.

Comment at: test/Driver/amdgpu-features.c:1
-// RUN: %clang -### -target amdgcn -x cl -S -emit-llvm -mcpu=kaveri -mamdgpu-debugger-abi=0.0 %s -o 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-MAMDGPU-DEBUGGER-ABI-0-0 %s
This should say `-o -`

Comment at: test/Driver/amdgpu-features.c:5
-// RUN: %clang -### -target amdgcn -x cl -S -emit-llvm -mcpu=kaveri -mamdgpu-debugger-abi=1.0 %s -o 2>&1 \
 // RUN:   | FileCheck --check-prefix=CHECK-MAMDGPU-DEBUGGER-ABI-1-0 %s

Comment at: test/Driver/split-debug.h:6-13
+// RUN: %clang -target x86_64-unknown-linux-gnu -gsplit-dwarf -c -fmodules -emit-module -fno-implicit-modules -fno-implicit-module-maps -### %s 2> %t
 // RUN: FileCheck -check-prefix=CHECK-NO-ACTIONS < %t %s
 // FIXME: This should fail using clang, except that the type of the output for
 // an object output with modules is given as clang::driver::types::TY_PCH
 // rather than TY_Object.
+// RUN: %clang -target x86_64-unknown-linux-gnu -gsplit-dwarf -c -fmodules -### %s 2> %t
These parts of the test don't make sense: the `-fmodule-format=obj` and `-emit-module` are `-cc1` options, so testing how the driver deals with them doesn't really make a lot of sense. I would suggest deleting the highlighted region of this test rather than making it test the same thing three times.


More information about the cfe-commits mailing list