[clang] [Clang][SYCL] Disable the tool timeout in sycl-system-arch.cpp (PR #238500)
via cfe-commits
cfe-commits at lists.llvm.org
Tue Nov 3 11:46:50 PST 2026
llvmorg-github-actions[bot] wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-clang
@llvm/pr-subscribers-clang-driver
Author: Srividya Sundaram (srividya-sundaram)
<details>
<summary>Changes</summary>
Disable the tool timeout in `sycl-system-arch.cpp`.
The `sycl-system-arch.cpp` test added in https://github.com/llvm/llvm-project/pull/229248 is failing on macOS pre-merge CI, e.g. https://github.com/llvm/llvm-project/actions/runs/41635790012/job/125916586515.
The test runs a fake `offload-arch` script, and clang waits at most 60 seconds for it. On the macOS CI machines, running such a script is very slow: the existing `nvptx-cuda-system-arch.c` and `openmp-system-arch.c` tests, which do the same, already take 3 to 4 minutes there. Since https://github.com/llvm/llvm-project/pull/229248 landed, this test failed with "Child timed out" in 29 of the 53 macOS runs that ran the clang tests.
This PR runs the test with `CLANG_TOOLCHAIN_PROGRAM_TIMEOUT=0`, so that clang waits for the fake tool without a limit.
---
Full diff: https://github.com/llvm/llvm-project/pull/238500.diff
1 Files Affected:
- (modified) clang/test/Driver/sycl-system-arch.cpp (+9-2)
``````````diff
diff --git a/clang/test/Driver/sycl-system-arch.cpp b/clang/test/Driver/sycl-system-arch.cpp
index b5c86ab760f04..fd129fac744f9 100644
--- a/clang/test/Driver/sycl-system-arch.cpp
+++ b/clang/test/Driver/sycl-system-arch.cpp
@@ -8,9 +8,15 @@
// RUN: chmod +x %t/offload_arch_sm_70_gfx906_gfx90a_xe_pvc
// RUN: chmod +x %t/offload_arch_sm_70_gfx90a
+/// By default, clang waits at most 60 seconds for offload-arch, and on a slow
+/// machine even starting the fake one can take longer. Set
+/// CLANG_TOOLCHAIN_PROGRAM_TIMEOUT=0 so that clang waits for it without a
+/// limit.
+
/// For SYCL offloading, --offload-arch=native asks offload-arch for Intel GPUs
/// only.
-// RUN: %clang -ccc-print-phases --target=x86_64-unknown-linux-gnu -fsycl \
+// RUN: env CLANG_TOOLCHAIN_PROGRAM_TIMEOUT=0 \
+// RUN: %clang -ccc-print-phases --target=x86_64-unknown-linux-gnu -fsycl \
// RUN: --offload-arch=native --offload-arch-tool=%t/offload_arch_sm_70_gfx906_gfx90a_xe_pvc \
// RUN: -c %s 2>&1 | FileCheck %s
// CHECK-NOT: "device-sycl ({{amdgpu|nvptx}}
@@ -19,7 +25,8 @@
/// Without an Intel GPU, SYCL reports that no GPU was found, as HIP and CUDA
/// do.
-// RUN: not %clang -### --target=x86_64-unknown-linux-gnu -fsycl \
+// RUN: env CLANG_TOOLCHAIN_PROGRAM_TIMEOUT=0 \
+// RUN: not %clang -### --target=x86_64-unknown-linux-gnu -fsycl \
// RUN: --offload-arch=native --offload-arch-tool=%t/offload_arch_sm_70_gfx90a \
// RUN: -c %s 2>&1 | FileCheck -check-prefix=NO-GPU %s
// NO-GPU: error: cannot determine sycl architecture: No GPU detected in the system
``````````
</details>
https://github.com/llvm/llvm-project/pull/238500
More information about the cfe-commits
mailing list