[PATCH] D46022: [OpenCL] Restrict various keywords in OpenCL C++ mode

Sven van Haastregt via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Thu Apr 26 09:26:11 PDT 2018


svenvh added inline comments.


================
Comment at: lib/Parse/ParseStmtAsm.cpp:696
+    return StmtError();
+  }
+
----------------
rjmccall wrote:
> You might consider parsing the statement normally and then just diagnosing after the fact, maybe in Sema.  You'd have to add the check in a couple different places, though.
Precisely the reason why I did it in the parser, otherwise we'd have to duplicate the check in multiple places.


================
Comment at: lib/Sema/DeclSpec.cpp:621
+    // OpenCL C++ 1.0 s2.9: the thread_local storage qualifier is not
+    // supported.
+    case TSCS_thread_local:
----------------
rjmccall wrote:
> svenvh wrote:
> > rjmccall wrote:
> > > Do you really care about the spelling of the specifier?  Presumably `__thread` (the GNU extension) and `_Thread_local` (the C11 feature) are unsupported; where are those diagnosed?
> > Fair enough, I have updated the patch to just reject any thread qualifier here.
> I meant it as a question, but I take it from this response that we don't actually diagnose these.
> 
> It might make more sense to diagnose this in `Sema::ActOnVariableDeclarator`, where we're already doing some checking about whether e.g. the target supports the feature.
We do actually diagnose the others in `Sema::ActOnVariableDeclarator` ("thread-local storage is not supported for the current target").  But your question made me realize there is no real need to differentiate here for the OpenCL C++ case.


================
Comment at: lib/Sema/SemaStmt.cpp:2791
+                     << "goto");
+  }
+
----------------
rjmccall wrote:
> Is this restriction really OpenCL C++-specific?  I mean, if that's what the language spec says, that's what it says, but I don't know what about OpenCL C++ would make `goto` unsupportable when it's supportable in ordinary OpenCL.
> 
> You should also add this check to ActOnIndirectGotoStmt.
Yes (oddly enough).  The OpenCL C++ 1.0 spec explicitly mentions `goto` in the Restrictions section on page 51.  The OpenCL C 1.1 / 2.0 specs ("ordinary OpenCL") do not restrict goto.

I'll add the check for indirect goto statements.


https://reviews.llvm.org/D46022





More information about the cfe-commits mailing list