[clang] [compiler-rt] Reapply " [XRay] Add support for instrumentation of DSOs on x86_64 (#90959)" (PR #113548)

Jordan Rupprecht via cfe-commits cfe-commits at lists.llvm.org
Mon Oct 28 20:04:12 PDT 2024


================
@@ -0,0 +1,17 @@
+// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fPIC -fxray-instrument -fxray-shared -c %s -o /dev/null 2>&1 | FileCheck %s
+// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fpic -fxray-instrument -fxray-shared -c %s -o /dev/null 2>&1 | FileCheck %s
+// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fxray-instrument -fxray-shared -c %s -o /dev/null 2>&1 | FileCheck %s
----------------
rupprecht wrote:

This test case fails for builds where the `CLANG_DEFAULT_PIE_ON_LINUX` cmake option is not the default.

If this test line is important, we could move it to a separate test file guarded by the `default-pie-on-linux` lit feature, i.e. like https://github.com/llvm/llvm-project/blob/main/clang/test/Driver/linux-default-pie.c.

However, I'd argue this case is not worth keeping. The two test lines above verify `-fxray-shared` work w/ pic, and the two below verify it doesn't work w/ no-pic. With the tests surrounding this in place, the only thing this verifies is that pic is the default (which isn't always true), and linux-default-pie.c already covers that behavior.

Any objection to just removing this line? Or should I move it to a separate file? (btw, not asking for a revert -- I'm just ignoring this test failure)

https://github.com/llvm/llvm-project/pull/113548


More information about the cfe-commits mailing list