build error: ../../Source/JavaScriptCore/llvm/library/LLVMExports.cpp: In function 'JSC::LLVMAPI* initializeAndGetJSCLLVMAPI(void (*)(const char*, ...) volatile, bool*)': ../../Source/JavaScriptCore/llvm/library/LLVMExports.cpp:132:17496: error: invalid conversion from 'LLVMOpaqueValue* (*)(LLVMBuilderRef, LLVMTypeRef, unsigned int, const char*) {aka LLVMOpaqueValue* (*)(LLVMOpaqueBuilder*, LLVMOpaqueType*, unsigned int, const char*)}' to 'LLVMOpaqueValue* (*)(LLVMBuilderRef, LLVMTypeRef, LLVMValueRef, unsigned int, const char*) {aka LLVMOpaqueValue* (*)(LLVMOpaqueBuilder*, LLVMOpaqueType*, LLVMOpaqueValue*, unsigned int, const char*)}' [-fpermissive]
Created attachment 264147 [details] Patch WIP patch, it fixed the build with LLVM 3.7, but we need a fix works with 3.6 and 3.7 too.
Created attachment 264150 [details] Patch
Created attachment 264204 [details] Patch EFL EWS was flakey, let's see again.
Comment on attachment 264204 [details] Patch I think Filip should review this.
(In reply to comment #4) > Comment on attachment 264204 [details] > Patch > > I think Filip should review this. Filip, could you possibly check it?
Ping?
Comment on attachment 264204 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=264204&action=review > Source/JavaScriptCore/llvm/LLVMAPI.h:50 > + > + // LLVMBuildLandingPad has different prototype with LLVM 3.7 and prior versions. > +#if LLVM_VERSION_MAJOR >= 4 || (LLVM_VERSION_MAJOR == 3 && LLVM_VERSION_MINOR >= 7) > + LLVMValueRef (*BuildLandingPad)(LLVMBuilderRef B, LLVMTypeRef Ty, unsigned NumClauses, const char *Name); > +#else > + LLVMValueRef (*BuildLandingPad)(LLVMBuilderRef B, LLVMTypeRef Ty, LLVMValueRef PersFn, unsigned NumClauses, const char *Name); > +#endif Let's just remove this function instead. We don't call it, and it's a mistake for it to be in the API struct.
Created attachment 265037 [details] Patch OK, let's remove it. Should we do a general cleanup for other functions too, or just leave them unchaged until we run into any issue?
(In reply to comment #8) > Created attachment 265037 [details] > Patch > > OK, let's remove it. Should we do a general cleanup for other functions too, > or just leave them unchaged until we run into any issue? Yes, I think that would be very useful!
Comment on attachment 265037 [details] Patch Clearing flags on attachment: 265037 Committed r192164: <http://trac.webkit.org/changeset/192164>
All reviewed patches have been landed. Closing bug.