[PATCH] D158231: [clang][test] Fix clang machine-function-split tests

[Han Shen via Phabricator via cfe-commits]

shenhan added inline comments.


================
Comment at: clang/test/CodeGen/fsplit-machine-functions.c:41
+
+// RUN: %clang -c -target arm-unknown-linux-gnueabi -fsplit-machine-functions %s \
+// RUN:     2>&1 | FileCheck -check-prefix=MFS5 %s
----------------
MaskRay wrote:
> `-target ` has been deprecated since Clang 3.4. Use `--target=`
> 
> 
> The warning test should be moved to `Driver/`.
> Whenever a warning can be placed either in Driver or in Frontend, we prefer Driver.
> -target  has been deprecated since Clang 3.4. Use --target=

Replaced with --target=arm-unknown-linux-gnueabi

> Whenever a warning can be placed either in Driver or in Frontend, we prefer Driver.

Just a little bit confused, you mentioned in [the other comment](https://reviews.llvm.org/D157750#inline-1530250)  that `Driver tests should not run the backend. Most driver commands should use -###.`
And the warning to be tested here is  issued from backend, so -### does not trigger it at all, so I moved the tests from Driver/ to here . Do you suggest move these back to Driver/?


================
Comment at: clang/test/Driver/fsplit-machine-functions.c:7
 
-// CHECK-OPT:       "-fsplit-machine-functions"
-// CHECK-NOOPT-NOT: "-fsplit-machine-functions"
-// CHECK-TRIPLE:    warning: -fsplit-machine-functions is not valid for arm
+// CHECK-OPT1:       "-fsplit-machine-functions"
+// CHECK-OPT2:       "-fsplit-machine-functions"
----------------
MaskRay wrote:
> if OPT1 is identical OPT2, we should use the same prefix.
> 
> If you want to test `-fprofile-use=`, test the extra CC1 option from the driver `-fprofile-use=`.
After a second thought, test of "fprofile-use" is irrelevant to fsplit-machine-functions flag, removed the test lines.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D158231/new/

https://reviews.llvm.org/D158231