[PATCH] D25363: [Sema] Store a SourceRange for multi-token builtin types

[Malcolm Parsons via cfe-commits]

malcolm.parsons added inline comments.


================
Comment at: include/clang/AST/TypeLoc.h:533
+    } else {
+      BuiltinRange.setBegin(std::min(Range.getBegin(), BuiltinRange.getBegin()));
+      BuiltinRange.setEnd(std::max(Range.getEnd(), BuiltinRange.getEnd()));
----------------
aaron.ballman wrote:
> malcolm.parsons wrote:
> > I suspect that using `min` and `max` on `SourceLocation`s is only valid if both locations are in the same file.
> > 
> > Doing this
> > long.h:
> > ```
> > long
> > ```
> > unsigned.cpp:
> > ```
> > unsigned
> > #include "long.h"
> > i;
> > ```
> > 
> > causes
> > clang-query> match typeLoc()
> > ...
> > clang-query: llvm/tools/clang/lib/Frontend/DiagnosticRenderer.cpp:273: clang::SourceLocation retrieveMacroLocation(clang::SourceLocation, clang::FileID, clang::FileID, const llvm::SmallVectorImpl<clang::FileID>&, bool, const clang::SourceManager*): Assertion `SM->getFileID(Loc) == MacroFileID' failed.
> > 
> > Is there a better way to combine `SourceRange`s, or should the final range be accumulated in DeclSpec.cpp?
> Hmm, that's a good point, but I'm not aware of a better utility. Perhaps @rsmith knows of one?
I get the same assertion failure from a clang-query without my changes if I match a `CompoundStmt` that starts and ends in different files.
So I don't think it's my fault.
It's also a very rare case.
OK to commit?


https://reviews.llvm.org/D25363