[Clang][C23] Crash when trying to use a `constexpr` variable that could not be initialized

[ianichitei]

I don't think I chose the best title, but I don't know how else to describe it, but the code is pretty straight forward: https://godbolt.org/z/1Tj8ba5bq

constexpr char c[] = { 'a' };
constexpr int i = c[1];
static_assert(i == c[0], "whatever");

When compiled with -std=c23 with clang 19.1.0 this happens:

<source>:2:19: error: constexpr variable 'i' must be initialized by a constant expression
    2 | constexpr int i = c[1];
      |                   ^~~~
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.
Stack dump:
0.	Program arguments: /opt/compiler-explorer/clang-19.1.0/bin/clang -gdwarf-4 -g -o /app/output.s -mllvm --x86-asm-syntax=intel -fno-verbose-asm -S --gcc-toolchain=/opt/compiler-explorer/gcc-13.2.0 -fcolor-diagnostics -fno-crash-diagnostics -std=c23 <source>
1.	<eof> parser at end of file
 #0 0x00000000036fdb08 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x36fdb08)
 #1 0x00000000036fbc74 llvm::sys::CleanupOnSignal(unsigned long) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x36fbc74)
 #2 0x0000000003647be8 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0
 #3 0x0000795267642520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)
 #4 0x00007952677aedcd (/lib/x86_64-linux-gnu/libc.so.6+0x1aedcd)
 #5 0x0000000007257afd clang::Expr::getIntegerConstantExpr(clang::ASTContext const&, clang::SourceLocation*) const (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x7257afd)
 #6 0x0000000006181049 AnalyzeImplicitConversions(clang::Sema&, clang::Expr*, clang::SourceLocation, bool) (.constprop.0) SemaChecking.cpp:0:0
 #7 0x00000000061833fa clang::Sema::CheckCompletedExpr(clang::Expr*, clang::SourceLocation, bool) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x61833fa)
 #8 0x00000000066204e4 clang::Sema::ActOnFinishFullExpr(clang::Expr*, clang::SourceLocation, bool, bool, bool) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x66204e4)
 #9 0x0000000006385d7d clang::Sema::BuildStaticAssertDeclaration(clang::SourceLocation, clang::Expr*, clang::Expr*, clang::SourceLocation, bool) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x6385d7d)
#10 0x0000000005f81e36 clang::Parser::ParseStaticAssertDeclaration(clang::SourceLocation&) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f81e36)
#11 0x0000000005f71028 clang::Parser::ParseDeclaration(clang::DeclaratorContext, clang::SourceLocation&, clang::ParsedAttributes&, clang::ParsedAttributes&, clang::SourceLocation*) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f71028)
#12 0x0000000005f27cfa clang::Parser::ParseExternalDeclaration(clang::ParsedAttributes&, clang::ParsedAttributes&, clang::ParsingDeclSpec*) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f27cfa)
#13 0x0000000005f29ac8 clang::Parser::ParseTopLevelDecl(clang::OpaquePtr<clang::DeclGroupRef>&, clang::Sema::ModuleImportState&) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f29ac8)
#14 0x0000000005f1877a clang::ParseAST(clang::Sema&, bool, bool) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f1877a)
#15 0x0000000003fce825 clang::CodeGenAction::ExecuteAction() (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3fce825)
#16 0x0000000004274201 clang::FrontendAction::Execute() (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x4274201)
#17 0x00000000041f699b clang::CompilerInstance::ExecuteAction(clang::FrontendAction&) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x41f699b)
#18 0x000000000435a843 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x435a843)
#19 0x0000000000c874dc cc1_main(llvm::ArrayRef<char const*>, char const*, void*) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0xc874dc)
#20 0x0000000000c8146d ExecuteCC1Tool(llvm::SmallVectorImpl<char const*>&, llvm::ToolContext const&) driver.cpp:0:0
#21 0x0000000004017309 void llvm::function_ref<void ()>::callback_fn<clang::driver::CC1Command::Execute(llvm::ArrayRef<std::optional<llvm::StringRef>>, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*, bool*) const::'lambda'()>(long) Job.cpp:0:0
#22 0x0000000003648013 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref<void ()>) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3648013)
#23 0x0000000004017529 clang::driver::CC1Command::Execute(llvm::ArrayRef<std::optional<llvm::StringRef>>, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char>>*, bool*) const (.part.0) Job.cpp:0:0
#24 0x0000000003fdd947 clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&, clang::driver::Command const*&, bool) const (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3fdd947)
#25 0x0000000003fde2fd clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&, llvm::SmallVectorImpl<std::pair<int, clang::driver::Command const*>>&, bool) const (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3fde2fd)
#26 0x0000000003fe7d4c clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&, llvm::SmallVectorImpl<std::pair<int, clang::driver::Command const*>>&) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3fe7d4c)
#27 0x0000000000c847f1 clang_main(int, char**, llvm::ToolContext const&) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0xc847f1)
#28 0x0000000000b5c2d4 main (/opt/compiler-explorer/clang-19.1.0/bin/clang+0xb5c2d4)
#29 0x0000795267629d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)
#30 0x0000795267629e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)
#31 0x0000000000c80f0e _start (/opt/compiler-explorer/clang-19.1.0/bin/clang+0xc80f0e)
clang: error: clang frontend command failed with exit code 139 (use -v to see invocation)
Compiler returned: 139

Weirdly enough, it looks like adding some other flags avoids the crash, for example with -std=c23 -O3: https://godbolt.org/z/41bGxqdh8 (it still crashes with -O3 -std=c23).

[llvmbot]

@llvm/issue-subscribers-clang-frontend

Author: Anichitei Ionel-Cristinel (ianichitei)

I don't think I chose the best title, but I don't know how else to describe it, but the code is pretty straight forward: https://godbolt.org/z/1Tj8ba5bq

constexpr char c[] = { 'a' };
constexpr int i = c[1];
static_assert(i == c[0], "whatever");

When compiled with -std=c23 with clang 19.1.0 this happens:

<source>:2:19: error: constexpr variable 'i' must be initialized by a constant expression
    2 | constexpr int i = c[1];
      |                   ^~~~
PLEASE submit a bug report to https://github.com/llvm/llvm-project/issues/ and include the crash backtrace, preprocessed source, and associated run script.
Stack dump:
0.	Program arguments: /opt/compiler-explorer/clang-19.1.0/bin/clang -gdwarf-4 -g -o /app/output.s -mllvm --x86-asm-syntax=intel -fno-verbose-asm -S --gcc-toolchain=/opt/compiler-explorer/gcc-13.2.0 -fcolor-diagnostics -fno-crash-diagnostics -std=c23 <source>
1.	<eof> parser at end of file
 #<!-- -->0 0x00000000036fdb08 llvm::sys::PrintStackTrace(llvm::raw_ostream&amp;, int) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x36fdb08)
 #<!-- -->1 0x00000000036fbc74 llvm::sys::CleanupOnSignal(unsigned long) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x36fbc74)
 #<!-- -->2 0x0000000003647be8 CrashRecoverySignalHandler(int) CrashRecoveryContext.cpp:0:0
 #<!-- -->3 0x0000795267642520 (/lib/x86_64-linux-gnu/libc.so.6+0x42520)
 #<!-- -->4 0x00007952677aedcd (/lib/x86_64-linux-gnu/libc.so.6+0x1aedcd)
 #<!-- -->5 0x0000000007257afd clang::Expr::getIntegerConstantExpr(clang::ASTContext const&amp;, clang::SourceLocation*) const (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x7257afd)
 #<!-- -->6 0x0000000006181049 AnalyzeImplicitConversions(clang::Sema&amp;, clang::Expr*, clang::SourceLocation, bool) (.constprop.0) SemaChecking.cpp:0:0
 #<!-- -->7 0x00000000061833fa clang::Sema::CheckCompletedExpr(clang::Expr*, clang::SourceLocation, bool) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x61833fa)
 #<!-- -->8 0x00000000066204e4 clang::Sema::ActOnFinishFullExpr(clang::Expr*, clang::SourceLocation, bool, bool, bool) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x66204e4)
 #<!-- -->9 0x0000000006385d7d clang::Sema::BuildStaticAssertDeclaration(clang::SourceLocation, clang::Expr*, clang::Expr*, clang::SourceLocation, bool) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x6385d7d)
#<!-- -->10 0x0000000005f81e36 clang::Parser::ParseStaticAssertDeclaration(clang::SourceLocation&amp;) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f81e36)
#<!-- -->11 0x0000000005f71028 clang::Parser::ParseDeclaration(clang::DeclaratorContext, clang::SourceLocation&amp;, clang::ParsedAttributes&amp;, clang::ParsedAttributes&amp;, clang::SourceLocation*) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f71028)
#<!-- -->12 0x0000000005f27cfa clang::Parser::ParseExternalDeclaration(clang::ParsedAttributes&amp;, clang::ParsedAttributes&amp;, clang::ParsingDeclSpec*) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f27cfa)
#<!-- -->13 0x0000000005f29ac8 clang::Parser::ParseTopLevelDecl(clang::OpaquePtr&lt;clang::DeclGroupRef&gt;&amp;, clang::Sema::ModuleImportState&amp;) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f29ac8)
#<!-- -->14 0x0000000005f1877a clang::ParseAST(clang::Sema&amp;, bool, bool) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x5f1877a)
#<!-- -->15 0x0000000003fce825 clang::CodeGenAction::ExecuteAction() (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3fce825)
#<!-- -->16 0x0000000004274201 clang::FrontendAction::Execute() (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x4274201)
#<!-- -->17 0x00000000041f699b clang::CompilerInstance::ExecuteAction(clang::FrontendAction&amp;) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x41f699b)
#<!-- -->18 0x000000000435a843 clang::ExecuteCompilerInvocation(clang::CompilerInstance*) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x435a843)
#<!-- -->19 0x0000000000c874dc cc1_main(llvm::ArrayRef&lt;char const*&gt;, char const*, void*) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0xc874dc)
#<!-- -->20 0x0000000000c8146d ExecuteCC1Tool(llvm::SmallVectorImpl&lt;char const*&gt;&amp;, llvm::ToolContext const&amp;) driver.cpp:0:0
#<!-- -->21 0x0000000004017309 void llvm::function_ref&lt;void ()&gt;::callback_fn&lt;clang::driver::CC1Command::Execute(llvm::ArrayRef&lt;std::optional&lt;llvm::StringRef&gt;&gt;, std::__cxx11::basic_string&lt;char, std::char_traits&lt;char&gt;, std::allocator&lt;char&gt;&gt;*, bool*) const::'lambda'()&gt;(long) Job.cpp:0:0
#<!-- -->22 0x0000000003648013 llvm::CrashRecoveryContext::RunSafely(llvm::function_ref&lt;void ()&gt;) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3648013)
#<!-- -->23 0x0000000004017529 clang::driver::CC1Command::Execute(llvm::ArrayRef&lt;std::optional&lt;llvm::StringRef&gt;&gt;, std::__cxx11::basic_string&lt;char, std::char_traits&lt;char&gt;, std::allocator&lt;char&gt;&gt;*, bool*) const (.part.0) Job.cpp:0:0
#<!-- -->24 0x0000000003fdd947 clang::driver::Compilation::ExecuteCommand(clang::driver::Command const&amp;, clang::driver::Command const*&amp;, bool) const (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3fdd947)
#<!-- -->25 0x0000000003fde2fd clang::driver::Compilation::ExecuteJobs(clang::driver::JobList const&amp;, llvm::SmallVectorImpl&lt;std::pair&lt;int, clang::driver::Command const*&gt;&gt;&amp;, bool) const (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3fde2fd)
#<!-- -->26 0x0000000003fe7d4c clang::driver::Driver::ExecuteCompilation(clang::driver::Compilation&amp;, llvm::SmallVectorImpl&lt;std::pair&lt;int, clang::driver::Command const*&gt;&gt;&amp;) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0x3fe7d4c)
#<!-- -->27 0x0000000000c847f1 clang_main(int, char**, llvm::ToolContext const&amp;) (/opt/compiler-explorer/clang-19.1.0/bin/clang+0xc847f1)
#<!-- -->28 0x0000000000b5c2d4 main (/opt/compiler-explorer/clang-19.1.0/bin/clang+0xb5c2d4)
#<!-- -->29 0x0000795267629d90 (/lib/x86_64-linux-gnu/libc.so.6+0x29d90)
#<!-- -->30 0x0000795267629e40 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x29e40)
#<!-- -->31 0x0000000000c80f0e _start (/opt/compiler-explorer/clang-19.1.0/bin/clang+0xc80f0e)
clang: error: clang frontend command failed with exit code 139 (use -v to see invocation)
Compiler returned: 139

Weirdly enough, it looks like adding some other flags avoids the crash, for example with -std=c23 -O3: https://godbolt.org/z/41bGxqdh8 (it still crashes with -O3 -std=c23).

[shafik]

Confirmed: https://godbolt.org/z/5PGzqeevv

It crashes unconditionally w/ assertion build though: https://godbolt.org/z/7Wb5snbEo

[shafik]

CC @Fznamznon @AaronBallman

[Fznamznon]

The problem here seems to be that due to variable being constexpr, isIntegerConstantExpr and CheckICE routines return true since for constexpr variables isUsableInConstantExpressions returns now true (which is correct, I think). We then end up in the codepath that assumes only valid integer constant expression here

llvm-project/clang/lib/AST/ExprConstant.cpp

Line 17277 in 7fa0d05

llvm_unreachable("ICE cannot be evaluated!");

It seems logical to not return true from isIntegerConstantExpr for invalid constexpr variables whose initializer is not a constant expression and avoid failing code path, like this:

@@ -2504,7 +2504,7 @@ bool VarDecl::isUsableInConstantExpressions(const ASTContext &Context) const {
   if (!DefVD->mightBeUsableInConstantExpressions(Context))
     return false;
   //   ... and its initializer is a constant initializer.
-  if (Context.getLangOpts().CPlusPlus && !DefVD->hasConstantInitialization())
+  if ((Context.getLangOpts().CPlusPlus || Context.getLangOpts().C23) && !DefVD->hasConstantInitialization())
     return false;
   // C++98 [expr.const]p1:
   //   An integral constant-expression can involve only [...] const variables

However hasConstantInitialization routine is problematic, since for C it without checking assumes that all globals have constant initialization (which is not the case for the variable in the example). However for constexpr variables we have the info in place, just like for C++ variables, so adding somewhat like

 bool VarDecl::hasConstantInitialization() const {
   // In C, all globals (and only globals) have constant initialization.
-  if (hasGlobalStorage() && !getASTContext().getLangOpts().CPlusPlus)
+  if (hasGlobalStorage() && !getASTContext().getLangOpts().CPlusPlus && !isConstexpr())
     return true;

In addition to the previous change in isUsableInConstantExpressions actually fixes the problem and passes testing.
But now it IMO seems like avoiding some global problem. In C++11 each constexpr evaluation is guarded with functions like EvaluateCPlusPlus11IntegralConstantExpr, isCXX11ConstantExpr etc. I don't really like this distinction, especially since we have C23 constexpr now. So, I wonder if it makes sense to follow the same codepath like EvaluateCPlusPlus11IntegralConstantExpr maybe with generalization of it for C++ and C23?
Or the easier way to fix the problem seems enough?

cc @AaronBallman @cor3ntin

[AaronBallman]

I don't really like this distinction, especially since we have C23 constexpr now. So, I wonder if it makes sense to follow the same codepath like EvaluateCPlusPlus11IntegralConstantExpr maybe with generalization of it for C++ and C23?

I think the larger refactoring makes more sense; ideally, the "high level" interfaces in the AST should be language agnostic because constexpr in C is intended to be compatible with the feature in C++, and we should make distinctions where necessary in the implementation-detail parts of the implementation (not to say we can't have helper functions that are C- or C++-specific where useful). However, if that looks to be a significant undertaking, the simpler approach is reasonable as a stop-gap. My fear with going with the easy approach is that each time we do that, the larger refactoring becomes a bit harder to implement.