New checks for fortified sprintf

[serge-sans-paille]

Summary

Implement a pessimistic evaluator of the minimal required size for a buffer based on the format string, and couple that with the fortified version to emit a warning when the buffer size is lower than the lower bound computed from the format string.

See the test cases for examples of warnings, and https://github.com/serge-sans-paille/llvm-project/pull/6/checks for the cross-platform validation.

Note: The lower bound could be improved, but I'd rather do that in an incremental commit, if that's okay with the reviewers.

Edit: thanks to @aaron.ballman interest, I've been able to support al(most|l) the scanf specification, ending up with a rather complete support.

[erik.pilkington]

Thanks for working on this! I think this would be a really useful diagnostic to have.

clang/lib/Sema/SemaChecking.cpp

392

nit: namespace {

580–581

Can you delete this comment now?

596

Optional<APSInt> might be a better way of representing this?

601

Can't you do this for Builtin::sprintf as well? The diagnostic would still be worth issuing when _FORTIFY_SOURCE is disabled.

605

Will this assert on: sprintf(buf, L"foo");? Not that that makes any sense, but we shouldn't crash.

612

I'm not sure why you're using min here, can you add a comment?

[serge-sans-paille]

Take remarks into account:

• support sprintf when the buffer has a known static size
• use llvm::Optional
• some extra minor tweaks.

[serge-sans-paille]

clang/lib/Sema/SemaChecking.cpp

580–581

I only deleted the one related to sprintf

605

Still need to check that.

[serge-sans-paille]

• Prevent assert upon wide string format
• More test cases

[serge-sans-paille]

clang/lib/Sema/SemaChecking.cpp

605

Checked and fixed!

[serge-sans-paille]

Support %% and %c

[serge-sans-paille]

@erik.pilkington up?

[serge-sans-paille]

@erik.pilkington up?

[erik.pilkington]

clang/lib/Sema/SemaChecking.cpp

621

Wait, does this actually return a smaller length if there is a null-terminator embedded in the string? It looks like Format->getString() returns a StringRef with embedded nul bytes and .size() is returning the length. You might have to do something like: FormatStrRef.find('0');. Can you add a test for this?

[serge-sans-paille]

Update size computation + test case involving null byte inside the format string.

[merge_guards_bot]

Unit tests: fail. 61975 tests passed, 1 failed and 783 were skipped.

failed: LLVM.Bindings/Go/go.test

clang-tidy: unknown.

clang-format: fail. Please format your changes with clang-format by running git-clang-format HEAD^ or applying this patch.

Build artifacts: diff.json, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[aaron.ballman]

clang/include/clang/Basic/DiagnosticSemaKinds.td

745–746

I'm fine with the current wording because it's consistent with the existing diagnostics, but I wish we would quantify the size with units in these diagnostics. (e.g., mention that this is measured in bytes as opposed to anything else).

clang/lib/Sema/SemaChecking.cpp

405–406

Naming nits: StartSpecifier and SpecifierLen. It looks like startSpecifier isn't used and the name can probably just be dropped.

436–437

Why is size incremented by 3 here instead of 2 as above with a leading 0x?

448

Guard against wraparound?

607

Spurious whitespace.

608

We can handle UTF-8 format strings even when they contain non-ASCII characters?

8382

Spurious whitespace change?

[serge-sans-paille]

Finer grain lower bound computations, and simpler control-flow for the checking function.

[merge_guards_bot]

Unit tests: pass. 62028 tests passed, 0 failed and 783 were skipped.

clang-tidy: unknown.

clang-format: fail. Please format your changes with clang-format by running git-clang-format HEAD^ or applying this patch.

Build artifacts: diff.json, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[serge-sans-paille]

More tests

[merge_guards_bot]

Unit tests: pass. 62043 tests passed, 0 failed and 783 were skipped.

clang-tidy: unknown.

clang-format: fail. Please format your changes with clang-format by running git-clang-format HEAD^ or applying this patch.

Build artifacts: diff.json, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[aaron.ballman]

(There are still some minor whitespace nits to resolve as well.)

clang/lib/Sema/SemaChecking.cpp

751

I think this assertion can be triggered in the case where no UsedSize was specified but then UsedSizeArg evaluates to 0, can't it? Or does something catch that case and diagnose?

[serge-sans-paille]

Remove useless assert

[serge-sans-paille]

In D71566#1832394, @aaron.ballman wrote:

(There are still some minor whitespace nits to resolve as well.)

Strange, everything is passed through clang-format-diff :-/

[merge_guards_bot]

Unit tests: pass. 62112 tests passed, 0 failed and 808 were skipped.

clang-tidy: pass.

clang-format: pass.

Build artifacts: diff.json, clang-tidy.txt, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

[aaron.ballman]

In D71566#1834472, @serge-sans-paille wrote:

In D71566#1832394, @aaron.ballman wrote:

(There are still some minor whitespace nits to resolve as well.)

Strange, everything is passed through clang-format-diff :-/

They may have been manually inserted by accident? It's newlines in a few places, I added phab review comments at them.

On the whole, I think this LGTM, assuming the requested test cases don't discover issues.

clang/test/Sema/warn-fortify-source.c

127

I'd like to see some additional tests for things like the + and flags, length modifiers like ll, escape characters, etc. Basically, we should be testing most of the conversion specifiers to verify our conservative length calculations.

[serge-sans-paille]

Remove extra new lines.
Add more test cases.

[merge_guards_bot]

Unit tests: pass. 62135 tests passed, 0 failed and 812 were skipped.

clang-tidy: pass.

clang-format: fail. Please format your changes with clang-format by running git-clang-format HEAD^ or applying this patch.

Build artifacts: diff.json, clang-tidy.txt, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

Pre-merge checks is in beta. Report issue. Please join beta or enable it for your project.

[serge-sans-paille]

clang-format update.

[merge_guards_bot]

Unit tests: fail. 62134 tests passed, 9 failed and 811 were skipped.

failed: LLVM.MC/AArch64/ete-sysregs.s
failed: LLVM.MC/AArch64/trace-regs.s
failed: LLVM.MC/Disassembler/AArch64/ete.txt
failed: LLVM.MC/Disassembler/AArch64/trace-regs.txt
failed: libc++.std/language_support/cmp/cmp_partialord/partialord.pass.cpp
failed: libc++.std/language_support/cmp/cmp_strongeq/cmp.strongeq.pass.cpp
failed: libc++.std/language_support/cmp/cmp_strongord/strongord.pass.cpp
failed: libc++.std/language_support/cmp/cmp_weakeq/cmp.weakeq.pass.cpp
failed: libc++.std/language_support/cmp/cmp_weakord/weakord.pass.cpp

clang-tidy: pass.

clang-format: pass.

Build artifacts: diff.json, clang-tidy.txt, clang-format.patch, CMakeCache.txt, console-log.txt, test-results.xml

Pre-merge checks is in beta. Report issue. Please join beta or enable it for your project.

[aaron.ballman]

Continues to LG, do you need me to commit on your behalf?

[serge-sans-paille]

In D71566#1839462, @aaron.ballman wrote:

Continues to LG, do you need me to commit on your behalf?

I was waiting for a last LG as I did some changes, thanks for all the quick iteration, I really enjoyed working with you on that patch! I'll submit it tomorrow morning.

clang/include/clang/Basic/DiagnosticSemaKinds.td

def warn_builtin_chk_overflow : Warning<
   InGroup<DiagGroup<"builtin-memcpy-chk-size">>;
 
 def warn_fortify_source_overflow
   : Warning<warn_builtin_chk_overflow.Text>, InGroup<FortifySource>;
 def warn_fortify_source_size_mismatch : Warning<
   "'%0' size argument is too large; destination buffer has size %1,"
   " but size argument is %2">, InGroup<FortifySource>;
 
+def warn_fortify_source_format_overflow : Warning<
+  "'%0' will always overflow; destination buffer has size %1,"
+  " but format string expands to at least %2">,
+  InGroup<FortifySource>;
+
+
 /// main()
 // static main() is not an error in C, just in C++.
 def warn_static_main : Warning<"'main' should not be declared static">,
     InGroup<Main>;
 def err_static_main : Error<"'main' is not allowed to be declared static">;
 def err_inline_main : Error<"'main' is not allowed to be declared inline">;
 def ext_variadic_main : ExtWarn<
   "'main' is not allowed to be declared variadic">, InGroup<Main>;

clang/lib/Sema/SemaChecking.cpp

static bool SemaBuiltinCallWithStaticChain(Sema &S, CallExpr *BuiltinCall) {
   BuiltinCall->setType(CE->getType());
   BuiltinCall->setValueKind(CE->getValueKind());
   BuiltinCall->setObjectKind(CE->getObjectKind());
   BuiltinCall->setCallee(Builtin);
   BuiltinCall->setArg(1, ChainResult.get());
 
   return false;
 }
 
+namespace {
+
+class EstimateSizeFormatHandler
+    : public analyze_format_string::FormatStringHandler {
+  size_t Size;
+
+public:
+  EstimateSizeFormatHandler(StringRef Format)
+      : Size(std::min(Format.find(0), Format.size()) +
+             1 /* null byte always written by sprintf */) {}
+
+  bool HandlePrintfSpecifier(const analyze_printf::PrintfSpecifier &FS,
+                             const char *, unsigned SpecifierLen) override {
+
+    const size_t FieldWidth = computeFieldWidth(FS);
+    const size_t Precision = computePrecision(FS);
+
+    // The actual format.
+    switch (FS.getConversionSpecifier().getKind()) {
+    // Just a char.
+    case analyze_format_string::ConversionSpecifier::cArg:
+    case analyze_format_string::ConversionSpecifier::CArg:
+      Size += std::max(FieldWidth, (size_t)1);
+      break;
+    // Just an integer.
+    case analyze_format_string::ConversionSpecifier::dArg:
+    case analyze_format_string::ConversionSpecifier::DArg:
+    case analyze_format_string::ConversionSpecifier::iArg:
+    case analyze_format_string::ConversionSpecifier::oArg:
+    case analyze_format_string::ConversionSpecifier::OArg:
+    case analyze_format_string::ConversionSpecifier::uArg:
+    case analyze_format_string::ConversionSpecifier::UArg:
+    case analyze_format_string::ConversionSpecifier::xArg:
+    case analyze_format_string::ConversionSpecifier::XArg:
+      Size += std::max(FieldWidth, Precision);
+      break;
+
+    // %g style conversion switches between %f or %e style dynamically.
+    // %f always takes less space, so default to it.
+    case analyze_format_string::ConversionSpecifier::gArg:
+    case analyze_format_string::ConversionSpecifier::GArg:
+
+    // Floating point number in the form '[+]ddd.ddd'.
+    case analyze_format_string::ConversionSpecifier::fArg:
+    case analyze_format_string::ConversionSpecifier::FArg:
+      Size += std::max(FieldWidth, 1 /* integer part */ +
+                                       (Precision ? 1 + Precision
+                                                  : 0) /* period + decimal */);
+      break;
+
+    // Floating point number in the form '[-]d.ddde[+-]dd'.
+    case analyze_format_string::ConversionSpecifier::eArg:
+    case analyze_format_string::ConversionSpecifier::EArg:
+      Size +=
+          std::max(FieldWidth,
+                   1 /* integer part */ +
+                       (Precision ? 1 + Precision : 0) /* period + decimal */ +
+                       1 /* e or E letter */ + 2 /* exponent */);
+      break;
+
+    // Floating point number in the form '[-]0xh.hhhhp±dd'.
+    case analyze_format_string::ConversionSpecifier::aArg:
+    case analyze_format_string::ConversionSpecifier::AArg:
+      Size +=
+          std::max(FieldWidth,
+                   2 /* 0x */ + 1 /* integer part */ +
+                       (Precision ? 1 + Precision : 0) /* period + decimal */ +
+                       1 /* p or P letter */ + 1 /* + or - */ + 1 /* value */);
+      break;
+
+    // Just a string.
+    case analyze_format_string::ConversionSpecifier::sArg:
+    case analyze_format_string::ConversionSpecifier::SArg:
+      Size += FieldWidth;
+      break;
+
+    // Just a pointer in the form '0xddd'.
+    case analyze_format_string::ConversionSpecifier::pArg:
+      Size += std::max(FieldWidth, 2 /* leading 0x */ + Precision);
+      break;
+
+    // A plain percent.
+    case analyze_format_string::ConversionSpecifier::PercentArg:
+      Size += 1;
+      break;
+
+    default:
+      break;
+    }
+
+    Size += FS.hasPlusPrefix() || FS.hasSpacePrefix();
+
+    if (FS.hasAlternativeForm()) {
+      switch (FS.getConversionSpecifier().getKind()) {
+      default:
+        break;
+      // Force a leading '0'.
+      case analyze_format_string::ConversionSpecifier::oArg:
+        Size += 1;
+        break;
+      // Force a leading '0x'.
+      case analyze_format_string::ConversionSpecifier::xArg:
+      case analyze_format_string::ConversionSpecifier::XArg:
+        Size += 2;
+        break;
+      // Force a period '.' before decimal, even if precision is 0.
+      case analyze_format_string::ConversionSpecifier::aArg:
+      case analyze_format_string::ConversionSpecifier::AArg:
+      case analyze_format_string::ConversionSpecifier::eArg:
+      case analyze_format_string::ConversionSpecifier::EArg:
+      case analyze_format_string::ConversionSpecifier::fArg:
+      case analyze_format_string::ConversionSpecifier::FArg:
+      case analyze_format_string::ConversionSpecifier::gArg:
+      case analyze_format_string::ConversionSpecifier::GArg:
+        Size += (Precision ? 0 : 1);
+        break;
+      }
+    }
+    assert(SpecifierLen <= Size && "no underflow");
+    Size -= SpecifierLen;
+    return true;
+  }
+
+  size_t getSizeLowerBound() const { return Size; }
+
+private:
+  static size_t computeFieldWidth(const analyze_printf::PrintfSpecifier &FS) {
+    const analyze_format_string::OptionalAmount &FW = FS.getFieldWidth();
+    size_t FieldWidth = 0;
+    if (FW.getHowSpecified() == analyze_format_string::OptionalAmount::Constant)
+      FieldWidth = FW.getConstantAmount();
+    return FieldWidth;
+  }
+
+  static size_t computePrecision(const analyze_printf::PrintfSpecifier &FS) {
+    const analyze_format_string::OptionalAmount &FW = FS.getPrecision();
+    size_t Precision = 0;
+
+    // See man 3 printf for default precision value based on the specifier.
+    switch (FW.getHowSpecified()) {
+    case analyze_format_string::OptionalAmount::NotSpecified:
+      switch (FS.getConversionSpecifier().getKind()) {
+      default:
+        break;
+      case analyze_format_string::ConversionSpecifier::dArg: // %d
+      case analyze_format_string::ConversionSpecifier::DArg: // %D
+      case analyze_format_string::ConversionSpecifier::iArg: // %i
+        Precision = 1;
+        break;
+      case analyze_format_string::ConversionSpecifier::oArg: // %d
+      case analyze_format_string::ConversionSpecifier::OArg: // %D
+      case analyze_format_string::ConversionSpecifier::uArg: // %d
+      case analyze_format_string::ConversionSpecifier::UArg: // %D
+      case analyze_format_string::ConversionSpecifier::xArg: // %d
+      case analyze_format_string::ConversionSpecifier::XArg: // %D
+        Precision = 1;
+        break;
+      case analyze_format_string::ConversionSpecifier::fArg: // %f
+      case analyze_format_string::ConversionSpecifier::FArg: // %F
+      case analyze_format_string::ConversionSpecifier::eArg: // %e
+      case analyze_format_string::ConversionSpecifier::EArg: // %E
+      case analyze_format_string::ConversionSpecifier::gArg: // %g
+      case analyze_format_string::ConversionSpecifier::GArg: // %G
+        Precision = 6;
+        break;
+      case analyze_format_string::ConversionSpecifier::pArg: // %d
+        Precision = 1;
+        break;
+      }
+      break;
+    case analyze_format_string::OptionalAmount::Constant:
+      Precision = FW.getConstantAmount();
+      break;
+    default:
+      break;
+    }
+    return Precision;
+  }
+};
+
+} // namespace
+
 /// Check a call to BuiltinID for buffer overflows. If BuiltinID is a
 /// __builtin_*_chk function, then use the object size argument specified in the
 /// source. Otherwise, infer the object size using __builtin_object_size.
 void Sema::checkFortifiedBuiltinMemoryFunction(FunctionDecl *FD,
                                                CallExpr *TheCall) {
   // FIXME: There are some more useful checks we could be doing here:
-  //  - Analyze the format string of sprintf to see how much of buffer is used.
   //  - Evaluate strlen of strcpy arguments, use as object size.
 
   if (TheCall->isValueDependent() || TheCall->isTypeDependent() ||
       isConstantEvaluated())
     return;
 
   unsigned BuiltinID = FD->getBuiltinID(/*ConsiderWrappers=*/true);
   if (!BuiltinID)
     return;
 
+  const TargetInfo &TI = getASTContext().getTargetInfo();
+  unsigned SizeTypeWidth = TI.getTypeWidth(TI.getSizeType());
+
   unsigned DiagID = 0;
   bool IsChkVariant = false;
+  Optional<llvm::APSInt> UsedSize;
   unsigned SizeIndex, ObjectIndex;
   switch (BuiltinID) {
   default:
     return;
+  case Builtin::BIsprintf:
+  case Builtin::BI__builtin___sprintf_chk: {
+    size_t FormatIndex = BuiltinID == Builtin::BIsprintf ? 1 : 3;
+    auto *FormatExpr = TheCall->getArg(FormatIndex)->IgnoreParenImpCasts();
+
+    if (auto *Format = dyn_cast<StringLiteral>(FormatExpr)) {
+
+      if (!Format->isAscii() && !Format->isUTF8())
+        return;
+
+      StringRef FormatStrRef = Format->getString();
+      EstimateSizeFormatHandler H(FormatStrRef);
+      const char *FormatBytes = FormatStrRef.data();
+      const ConstantArrayType *T =
+          Context.getAsConstantArrayType(Format->getType());
+      assert(T && "String literal not of constant array type!");
+      size_t TypeSize = T->getSize().getZExtValue();
+
+      // In case there's a null byte somewhere.
+      size_t StrLen =
+          std::min(std::max(TypeSize, size_t(1)) - 1, FormatStrRef.find(0));
+      if (!analyze_format_string::ParsePrintfString(
+              H, FormatBytes, FormatBytes + StrLen, getLangOpts(),
+              Context.getTargetInfo(), false)) {
+        DiagID = diag::warn_fortify_source_format_overflow;
+        UsedSize = llvm::APSInt::getUnsigned(H.getSizeLowerBound())
+                       .extOrTrunc(SizeTypeWidth);
+        if (BuiltinID == Builtin::BI__builtin___sprintf_chk) {
+          IsChkVariant = true;
+          ObjectIndex = 2;
+        } else {
+          IsChkVariant = false;
+          ObjectIndex = 0;
+        }
+        break;
+      }
+    }
+    return;
+  }
   case Builtin::BI__builtin___memcpy_chk:
   case Builtin::BI__builtin___memmove_chk:
   case Builtin::BI__builtin___memset_chk:
   case Builtin::BI__builtin___strlcat_chk:
   case Builtin::BI__builtin___strlcpy_chk:
   case Builtin::BI__builtin___strncat_chk:
   case Builtin::BI__builtin___strncpy_chk:
   case Builtin::BI__builtin___stpncpy_chk:
if (const auto *POS =
             FD->getParamDecl(ObjectIndex)->getAttr<PassObjectSizeAttr>())
       BOSType = POS->getType();
 
     Expr *ObjArg = TheCall->getArg(ObjectIndex);
     uint64_t Result;
     if (!ObjArg->tryEvaluateObjectSize(Result, getASTContext(), BOSType))
       return;
     // Get the object size in the target's size_t width.
-    const TargetInfo &TI = getASTContext().getTargetInfo();
-    unsigned SizeTypeWidth = TI.getTypeWidth(TI.getSizeType());
     ObjectSize = llvm::APSInt::getUnsigned(Result).extOrTrunc(SizeTypeWidth);
   }
 
   // Evaluate the number of bytes of the object that this call will use.
+  if (!UsedSize) {
     Expr::EvalResult Result;
     Expr *UsedSizeArg = TheCall->getArg(SizeIndex);
     if (!UsedSizeArg->EvaluateAsInt(Result, getASTContext()))
       return;
-  llvm::APSInt UsedSize = Result.Val.getInt();
+    UsedSize = Result.Val.getInt().extOrTrunc(SizeTypeWidth);
+  }
 
-  if (UsedSize.ule(ObjectSize))
+  if (UsedSize.getValue().ule(ObjectSize))
     return;
 
   StringRef FunctionName = getASTContext().BuiltinInfo.getName(BuiltinID);
   // Skim off the details of whichever builtin was called to produce a better
   // diagnostic, as it's unlikley that the user wrote the __builtin explicitly.
   if (IsChkVariant) {
     FunctionName = FunctionName.drop_front(std::strlen("__builtin___"));
     FunctionName = FunctionName.drop_back(std::strlen("_chk"));
   } else if (FunctionName.startswith("__builtin_")) {
     FunctionName = FunctionName.drop_front(std::strlen("__builtin_"));
   }
 
   DiagRuntimeBehavior(TheCall->getBeginLoc(), TheCall,
                       PDiag(DiagID)
                           << FunctionName << ObjectSize.toString(/*Radix=*/10)
-                          << UsedSize.toString(/*Radix=*/10));
+                          << UsedSize.getValue().toString(/*Radix=*/10));
 }
 
 static bool SemaBuiltinSEHScopeCheck(Sema &SemaRef, CallExpr *TheCall,
                                      Scope::ScopeFlags NeededScopeFlags,
                                      unsigned DiagID) {
   // Scopes aren't available during instantiation. Fortunately, builtin
   // functions cannot be template args so they cannot be formed through template
   // instantiation. Therefore checking once during the parse is sufficient.
CheckPrintfHandler H(
         HasVAListArg, Args, format_idx, inFunctionCall, CallType,
         CheckedVarArgs, UncoveredArg);
 
     if (!analyze_format_string::ParsePrintfString(H, Str, Str + StrLen,
                                                   S.getLangOpts(),
                                                   S.Context.getTargetInfo(),
                                             Type == Sema::FST_FreeBSDKPrintf))
       H.DoneProcessing();
   } else if (Type == Sema::FST_Scanf) {
     CheckScanfHandler H(S, FExpr, OrigFormatExpr, Type, firstDataArg,
                         numDataArgs, Str, HasVAListArg, Args, format_idx,
                         inFunctionCall, CallType, CheckedVarArgs, UncoveredArg);
 
     if (!analyze_format_string::ParseScanfString(H, Str, Str + StrLen,
                                                  S.getLangOpts(),
                                                  S.Context.getTargetInfo()))
       H.DoneProcessing();

clang/test/Sema/warn-fortify-source.c

 // RUN: %clang_cc1 -triple x86_64-apple-macosx10.14.0 %s -verify
 // RUN: %clang_cc1 -triple x86_64-apple-macosx10.14.0 %s -verify -DUSE_PASS_OBJECT_SIZE
 // RUN: %clang_cc1 -triple x86_64-apple-macosx10.14.0 %s -verify -DUSE_BUILTINS
 // RUN: %clang_cc1 -xc++ -triple x86_64-apple-macosx10.14.0 %s -verify
 // RUN: %clang_cc1 -xc++ -triple x86_64-apple-macosx10.14.0 %s -verify -DUSE_PASS_OBJECT_SIZE
 // RUN: %clang_cc1 -xc++ -triple x86_64-apple-macosx10.14.0 %s -verify -DUSE_BUILTINS
 
 typedef unsigned long size_t;
 
 #ifdef __cplusplus
 extern "C" {
 #endif
 
+extern int sprintf(char *str, const char *format, ...);
+
 #if defined(USE_PASS_OBJECT_SIZE)
 void *memcpy(void *dst, const void *src, size_t c);
 static void *memcpy(void *dst __attribute__((pass_object_size(1))), const void *src, size_t c) __attribute__((overloadable)) __asm__("merp");
 static void *memcpy(void *const dst __attribute__((pass_object_size(1))), const void *src, size_t c) __attribute__((overloadable)) {
   return 0;
 }
 #elif defined(USE_BUILTINS)
 #define memcpy(x,y,z) __builtin_memcpy(x,y,z)

 
 void call_vsnprintf() {
   char buf[10];
   __builtin_va_list list;
   __builtin_vsnprintf(buf, 10, "merp", list);
   __builtin_vsnprintf(buf, 11, "merp", list); // expected-warning {{'vsnprintf' size argument is too large; destination buffer has size 10, but size argument is 11}}
 }
 
+void call_sprintf_chk(char *buf) {
+  __builtin___sprintf_chk(buf, 1, 6, "hell\n");
+  __builtin___sprintf_chk(buf, 1, 5, "hell\n");     // expected-warning {{'sprintf' will always overflow; destination buffer has size 5, but format string expands to at least 6}}
+  __builtin___sprintf_chk(buf, 1, 6, "hell\0 boy"); // expected-warning {{format string contains '\0' within the string body}}
+  __builtin___sprintf_chk(buf, 1, 2, "hell\0 boy"); // expected-warning {{format string contains '\0' within the string body}}
+  // expected-warning@-1 {{'sprintf' will always overflow; destination buffer has size 2, but format string expands to at least 5}}
+  __builtin___sprintf_chk(buf, 1, 6, "hello");
+  __builtin___sprintf_chk(buf, 1, 5, "hello"); // expected-warning {{'sprintf' will always overflow; destination buffer has size 5, but format string expands to at least 6}}
+  __builtin___sprintf_chk(buf, 1, 2, "%c", '9');
+  __builtin___sprintf_chk(buf, 1, 1, "%c", '9'); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%d", 9);
+  __builtin___sprintf_chk(buf, 1, 1, "%d", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%i", 9);
+  __builtin___sprintf_chk(buf, 1, 1, "%i", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%o", 9);
+  __builtin___sprintf_chk(buf, 1, 1, "%o", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%u", 9);
+  __builtin___sprintf_chk(buf, 1, 1, "%u", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%x", 9);
+  __builtin___sprintf_chk(buf, 1, 1, "%x", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%X", 9);
+  __builtin___sprintf_chk(buf, 1, 1, "%X", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%hhd", (char)9);
+  __builtin___sprintf_chk(buf, 1, 1, "%hhd", (char)9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%hd", (short)9);
+  __builtin___sprintf_chk(buf, 1, 1, "%hd", (short)9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%ld", 9l);
+  __builtin___sprintf_chk(buf, 1, 1, "%ld", 9l); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%lld", 9ll);
+  __builtin___sprintf_chk(buf, 1, 1, "%lld", 9ll); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 2, "%%");
+  __builtin___sprintf_chk(buf, 1, 1, "%%"); // expected-warning {{'sprintf' will always overflow; destination buffer has size 1, but format string expands to at least 2}}
+  __builtin___sprintf_chk(buf, 1, 4, "%#x", 9);
+  __builtin___sprintf_chk(buf, 1, 3, "%#x", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 3, but format string expands to at least 4}}
+  __builtin___sprintf_chk(buf, 1, 4, "%p", (void *)9);
+  __builtin___sprintf_chk(buf, 1, 3, "%p", (void *)9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 3, but format string expands to at least 4}}
+  __builtin___sprintf_chk(buf, 1, 3, "%+d", 9);
+  __builtin___sprintf_chk(buf, 1, 2, "%+d", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 2, but format string expands to at least 3}}
+  __builtin___sprintf_chk(buf, 1, 3, "% i", 9);
+  __builtin___sprintf_chk(buf, 1, 2, "% i", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 2, but format string expands to at least 3}}
+  __builtin___sprintf_chk(buf, 1, 6, "%5d", 9);
+  __builtin___sprintf_chk(buf, 1, 5, "%5d", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 5, but format string expands to at least 6}}
+  __builtin___sprintf_chk(buf, 1, 9, "%f", 9.f);
+  __builtin___sprintf_chk(buf, 1, 8, "%f", 9.f); // expected-warning {{'sprintf' will always overflow; destination buffer has size 8, but format string expands to at least 9}}
+  __builtin___sprintf_chk(buf, 1, 9, "%Lf", (long double)9.);
+  __builtin___sprintf_chk(buf, 1, 8, "%Lf", (long double)9.); // expected-warning {{'sprintf' will always overflow; destination buffer has size 8, but format string expands to at least 9}}
+  __builtin___sprintf_chk(buf, 1, 10, "%+f", 9.f);
+  __builtin___sprintf_chk(buf, 1, 9, "%+f", 9.f); // expected-warning {{'sprintf' will always overflow; destination buffer has size 9, but format string expands to at least 10}}
+  __builtin___sprintf_chk(buf, 1, 12, "%e", 9.f);
+  __builtin___sprintf_chk(buf, 1, 11, "%e", 9.f); // expected-warning {{'sprintf' will always overflow; destination buffer has size 11, but format string expands to at least 12}}
+}
+
+void call_sprintf() {
+  char buf[6];
+  sprintf(buf, "hell\0 boy"); // expected-warning {{format string contains '\0' within the string body}}
+  sprintf(buf, "hello b\0y"); // expected-warning {{format string contains '\0' within the string body}}
+  // expected-warning@-1 {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 8}}
+  sprintf(buf, "hello");
+  sprintf(buf, "hello!"); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "1234%%");
+  sprintf(buf, "12345%%"); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "1234%c", '9');
+  sprintf(buf, "12345%c", '9'); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "1234%d", 9);
+  sprintf(buf, "12345%d", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "1234%lld", 9ll);
+  sprintf(buf, "12345%lld", 9ll); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "12%#x", 9);
+  sprintf(buf, "123%#x", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "12%p", (void *)9);
+  sprintf(buf, "123%p", (void *)9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "123%+d", 9);
+  sprintf(buf, "1234%+d", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "123% i", 9);
+  sprintf(buf, "1234% i", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "%5d", 9);
+  sprintf(buf, "1%5d", 9); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "%.3f", 9.f);
+  sprintf(buf, "5%.3f", 9.f); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "%+.2f", 9.f);
+  sprintf(buf, "%+.3f", 9.f); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 7}}
+  sprintf(buf, "%.0e", 9.f);
+  sprintf(buf, "5%.1e", 9.f); // expected-warning {{'sprintf' will always overflow; destination buffer has size 6, but format string expands to at least 8}}
+}
+
 #ifdef __cplusplus
 template <class> struct S {
   void mf() const {
     __builtin_memset(const_cast<char *>(mv), 0, 0);
   }
 
   char mv[10];
 };