search for: llvmgetbitcodemoduleprovid

...;> prevent exceptions being passed on to C callers? > > LLVM doesn't use exceptions in anything that has C bindings, so the > catch isn't needed. It does use 'new' (e.g. LLVMCreateModuleProviderForExistingModule), and also calls C++ functions which in turn use new (e.g. LLVMGetBitcodeModuleProvider -> llvm::getBitcodeModuleProvider), so std::bad_alloc is a possibility (a quick search didn't turn up any "set_new_handler" also). There might also be C++ client code (e.g. overridden virtual functions in custom passes) which might (unintentionally) throw exceptions. Regards, -...

...geting LLVM 2.5 to the top of tree, but I'm running into a linking problem I can't figure out. Here's how I'm invoking the JIT in 2.5. LLVMInitializeNativeTarget(); <-- for the updated version LLVMCreateMemoryBufferWithContentsOfFile("code.bc", &buffer, &err); LLVMGetBitcodeModuleProvider(buffer, &provider, &err); LLVMCreateJITCompiler(&engine, provider, 1, &err); After that LLVMFindFunction and RunFunction work wonderfully on 2.5. The TOT version errors out on the CreateJITCompiler call with a "JIT has not been linked in" message. I'm compiling it...

...tree, but I'm running into a linking problem I can't figure out. > > Here's how I'm invoking the JIT in 2.5. > > LLVMInitializeNativeTarget(); <-- for the updated version > LLVMCreateMemoryBufferWithContentsOfFile("code.bc", &buffer, &err); > LLVMGetBitcodeModuleProvider(buffer, &provider, &err); > LLVMCreateJITCompiler(&engine, provider, 1, &err); > > After that LLVMFindFunction and RunFunction work wonderfully on 2.5. The TOT version errors out on the CreateJITCompiler call with a "JIT has not been linked in" message. >...

Hi, Just wondering -- shouldn't all (C linkage) functions exposed by LLVM-C (and written in C++) be catching std::exception (or "...") to prevent exceptions being passed on to C callers? [OT: Does clang warn about throw statements from within "extern C" functions?] Thanks & Regards, -Mahadevan.

On Aug 3, 2008, at 7:54 AM, Mahadevan R wrote: > Hi, > > Just wondering -- shouldn't all (C linkage) functions exposed by > LLVM-C (and written in C++) be catching std::exception (or "...") to > prevent exceptions being passed on to C callers? LLVM doesn't use exceptions in anything that has C bindings, so the catch isn't needed. > [OT: Does clang warn