[PATCH] D55534: [AST] Store "UsesADL" information in CallExpr.

[Eric Fiselier via Phabricator via cfe-commits]

EricWF marked 14 inline comments as done.
EricWF added a comment.

Address review comments. Fixes incoming.



================
Comment at: include/clang/AST/ExprCXX.h:218
+                     SourceLocation RP, unsigned MinNumArgs = 0,
+                     bool UsesADL = false)
       : CallExpr(C, CUDAKernelCallExprClass, fn, Config, args, t, VK, RP,
----------------
fowles wrote:
> Can CUDAKernelCalls ever use ADL?
Yeah, after further investigation, I don't think so. I'll revert this change.


================
Comment at: include/clang/AST/ExprCXX.h:499
+      : CallExpr(C, UserDefinedLiteralClass, Fn, Args, T, VK, LitEndLoc,
+                 /*MinNumArgs=*/0, /*UsesADL=*/false),
         UDSuffixLoc(SuffixLoc) {}
----------------
fowles wrote:
> Same question, I assume user defined literals can only be called as string suffixes and thus never get ADL
Right. Their argument is a literal of some sort. 


================
Comment at: lib/Sema/SemaOverload.cpp:8946
+                                   ExplicitTemplateArgs, Args, CandidateSet,
+                                   /*SuppressUserConversions=*/false,
+                                   PartialOverloading, /*IsADLCandidate=*/true);
----------------
fowles wrote:
> should this fix be in a separate change?
I'll commit the fix as a separate patch.


================
Comment at: lib/Serialization/ASTReaderStmt.cpp:735
   unsigned NumArgs = Record.readInt();
+  bool UsesADL = Record.readInt();
   assert((NumArgs == E->getNumArgs()) && "Wrong NumArgs!");
----------------
fowles wrote:
> readInt into a bool?
Yeah. You serialize and serialize all integer types using `readInt()`


================
Comment at: unittests/ASTMatchers/ASTMatchersNodeTest.cpp:204
+  StatementMatcher ADLMatch = callExpr(usesADL());
+  auto NS_Str = R"DELIM(
+  namespace NS {
----------------
fowles wrote:
> if you use cc or cpp as your delimiter, I think clang-format will recurse correctly
That's super cool. Thanks!


Repository:
  rC Clang

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D55534/new/

https://reviews.llvm.org/D55534