[llvm] r333147 - [ORC] Add findSymbolIn() wrapper to C bindings.

[Andres Freund via llvm-commits]

Hi,

On 2018-05-23 21:08:19 -0700, Andres Freund wrote:
> On 2018-05-23 20:24:16 -0700, Andres Freund wrote:
> > On 2018-05-24 01:01:42 -0000, Andres Freund via llvm-commits wrote:
> > The green.lab.llvm.org builder seems to be OSX - I can't for the heck of
> > me see what was OS specific in that change however.
> > 
> > I've organized remote access to an OSX laptop (using linux myself), and
> > will build LLVM there to see whether I can reproduce the problem. On
> > linux the tests pass, even under valgrind.
> 
> That helped. As far as I can tell the bug is in pre-existing code that
> this commit exposed. Compare:
> 
>   JITSymbol findSymbol(const std::string &Name,
>                                  bool ExportedSymbolsOnly) {
>     if (auto Sym = IndirectStubsMgr->findStub(Name, ExportedSymbolsOnly))
>       return Sym;
>     return CODLayer.findSymbol(mangle(Name), ExportedSymbolsOnly);
>   }
> 
> with
> 
>   JITSymbol findSymbolIn(orc::VModuleKey K, const std::string &Name,
>                          bool ExportedSymbolsOnly) {
>     assert(KeyLayers.count(K) && "looking up symbol in unknown module");
>     return KeyLayers[K]->findSymbolIn(K, Name, ExportedSymbolsOnly);
>   }
> 
> Note that the second function doesn't do a mangle(Name). On OSX the
> actual symbol will have a '_' prefix, but without mangling the looked up
> function won't.
> 
> I'll open a review about this.

Submitted as https://reviews.llvm.org/D47308

Seems like the fairly obvious fix to me, but I'm not clear enough on the
review policy to commit this without review.

Greetings,

Andres Freund