[clang][Driver][AVR] Reject c/c++ compilation for avr1 devices

[benshi001]

avr-gcc also rejects since these devices has no SRAM.

Fixes #96881

[llvmbot]

@llvm/pr-subscribers-clang

Author: Ben Shi (benshi001)

Changes

avr-gcc also rejects since these devices has no SRAM.

---

Full diff: https://github.com/llvm/llvm-project/pull/111798.diff

2 Files Affected:

• (modified) clang/lib/Driver/ToolChains/AVR.cpp (+8)
• (modified) clang/test/Driver/avr-mmcu.c (+2-3)

diff --git a/clang/lib/Driver/ToolChains/AVR.cpp b/clang/lib/Driver/ToolChains/AVR.cpp
index bb5c0e6db9978e..83a756b07056ca 100644
--- a/clang/lib/Driver/ToolChains/AVR.cpp
+++ b/clang/lib/Driver/ToolChains/AVR.cpp
@@ -400,6 +400,14 @@ void AVRToolChain::AddClangSystemIncludeArgs(const ArgList &DriverArgs,
 void AVRToolChain::addClangTargetOptions(
     const llvm::opt::ArgList &DriverArgs, llvm::opt::ArgStringList &CC1Args,
     Action::OffloadKind DeviceOffloadKind) const {
+  // Reject C/C++ compilation for avr1 devices.
+  const Driver &D = getDriver();
+  std::string CPU = getCPUName(D, DriverArgs, getTriple());
+  std::optional<StringRef> FamilyName = GetMCUFamilyName(CPU);
+  if (CPU == "avr1" || (FamilyName && FamilyName->compare("avr1") == 0))
+    D.Diag(diag::err_drv_opt_unsupported_input_type) << "avr1"
+                                                     << "c/c++";
+
   // By default, use `.ctors` (not `.init_array`), as required by libgcc, which
   // runs constructors/destructors on AVR.
   if (!DriverArgs.hasFlag(options::OPT_fuse_init_array,
diff --git a/clang/test/Driver/avr-mmcu.c b/clang/test/Driver/avr-mmcu.c
index 6c13ab5b68d3b9..e354917ae9ab1d 100644
--- a/clang/test/Driver/avr-mmcu.c
+++ b/clang/test/Driver/avr-mmcu.c
@@ -1,8 +1,7 @@
 // A test for the propagation of the -mmcu option to -cc1 and -cc1as
 
-// RUN: %clang -### --target=avr -mmcu=attiny11 -save-temps %s 2>&1 | FileCheck -check-prefix=CHECK0 %s
-// CHECK0: "-cc1" {{.*}} "-target-cpu" "attiny11"
-// CHECK0: "-cc1as" {{.*}} "-target-cpu" "attiny11"
+// RUN: not %clang -### --target=avr -mmcu=attiny11 %s 2>&1 | FileCheck -check-prefix=CHECK0 %s
+// CHECK0: error: 'avr1' invalid for input of type c/c++
 
 // RUN: %clang -### --target=avr -mmcu=at90s2313 -save-temps %s 2>&1 | FileCheck -check-prefix=CHECK1 %s
 // CHECK1: "-cc1" {{.*}} "-target-cpu" "at90s2313"

[llvmbot]

@llvm/pr-subscribers-clang-driver

Author: Ben Shi (benshi001)

Changes

avr-gcc also rejects since these devices has no SRAM.

---

Full diff: https://github.com/llvm/llvm-project/pull/111798.diff

2 Files Affected:

• (modified) clang/lib/Driver/ToolChains/AVR.cpp (+8)
• (modified) clang/test/Driver/avr-mmcu.c (+2-3)

diff --git a/clang/lib/Driver/ToolChains/AVR.cpp b/clang/lib/Driver/ToolChains/AVR.cpp
index bb5c0e6db9978e..83a756b07056ca 100644
--- a/clang/lib/Driver/ToolChains/AVR.cpp
+++ b/clang/lib/Driver/ToolChains/AVR.cpp
@@ -400,6 +400,14 @@ void AVRToolChain::AddClangSystemIncludeArgs(const ArgList &DriverArgs,
 void AVRToolChain::addClangTargetOptions(
     const llvm::opt::ArgList &DriverArgs, llvm::opt::ArgStringList &CC1Args,
     Action::OffloadKind DeviceOffloadKind) const {
+  // Reject C/C++ compilation for avr1 devices.
+  const Driver &D = getDriver();
+  std::string CPU = getCPUName(D, DriverArgs, getTriple());
+  std::optional<StringRef> FamilyName = GetMCUFamilyName(CPU);
+  if (CPU == "avr1" || (FamilyName && FamilyName->compare("avr1") == 0))
+    D.Diag(diag::err_drv_opt_unsupported_input_type) << "avr1"
+                                                     << "c/c++";
+
   // By default, use `.ctors` (not `.init_array`), as required by libgcc, which
   // runs constructors/destructors on AVR.
   if (!DriverArgs.hasFlag(options::OPT_fuse_init_array,
diff --git a/clang/test/Driver/avr-mmcu.c b/clang/test/Driver/avr-mmcu.c
index 6c13ab5b68d3b9..e354917ae9ab1d 100644
--- a/clang/test/Driver/avr-mmcu.c
+++ b/clang/test/Driver/avr-mmcu.c
@@ -1,8 +1,7 @@
 // A test for the propagation of the -mmcu option to -cc1 and -cc1as
 
-// RUN: %clang -### --target=avr -mmcu=attiny11 -save-temps %s 2>&1 | FileCheck -check-prefix=CHECK0 %s
-// CHECK0: "-cc1" {{.*}} "-target-cpu" "attiny11"
-// CHECK0: "-cc1as" {{.*}} "-target-cpu" "attiny11"
+// RUN: not %clang -### --target=avr -mmcu=attiny11 %s 2>&1 | FileCheck -check-prefix=CHECK0 %s
+// CHECK0: error: 'avr1' invalid for input of type c/c++
 
 // RUN: %clang -### --target=avr -mmcu=at90s2313 -save-temps %s 2>&1 | FileCheck -check-prefix=CHECK1 %s
 // CHECK1: "-cc1" {{.*}} "-target-cpu" "at90s2313"

[DavidSpickett]

Thanks for getting to this I had forgotten I raised that issue.

You should move your commit's message into the PR description, as the PR description is what's used for the final commit when we merge this.

[DavidSpickett]

�_bk;t=1728546187607�Failed Tests (1):
�_bk;t=1728546187607�  Clang :: Driver/hip-partial-link.hip

And the CI test failure is this, the reporting is not great because we run check- targets one after another. This test was already failing on main.

If it fails again download the logs and search for "Failed<2-3 spaces>" and you'll see any results summaries where something failed. I plan to work on combining the results so this isn't a problem in the future.

[benshi001]

Thanks for getting to this I had forgotten I raised that issue.

You should move your commit's message into the PR description, as the PR description is what's used for the final commit when we merge this.

Thanks. I have update my PR description according to my commit message.

[benshi001]

�_bk;t=1728546187607�Failed Tests (1):
�_bk;t=1728546187607�  Clang :: Driver/hip-partial-link.hip

And the CI test failure is this, the reporting is not great because we run check- targets one after another. This test was already failing on main.

If it fails again download the logs and search for "Failed<2-3 spaces>" and you'll see any results summaries where something failed. I plan to work on combining the results so this isn't a problem in the future.

Thanks. I am also thinking the HIP failure is not related to my clang-driver-AVR work.

[DavidSpickett]

Please add a release note in https://clang.llvm.org/docs/ReleaseNotes.html#avr-support (clang/docs/ReleaseNotes.rst).

Otherwise, this looks good to me.

[benshi001]

Please add a release note in https://clang.llvm.org/docs/ReleaseNotes.html#avr-support (clang/docs/ReleaseNotes.rst).

Otherwise, this looks good to me.

The note is added. Thanks!

[DavidSpickett]

LGTM with the wording fixed.