lld-link: Also demangle undefined dllimported symbols.

[thakis]

Summary

dllimported symbols go through an import stub that's called __imp_ followed by the name the stub points to. Make that work.

Follow-up to https://reviews.llvm.org/D52104

[thakis]

fix case on variable

[ruiu]

LGTM

Common/Strings.cpp

41 ↗

(On Diff #165672)

Is "imp_" name mangling really only for C++ symbol names? I wonder if it's wrong to demangle imp_f as "__declspec(dllimport) f".

[ruiu]

Oh sorry I didn't intend to LGTM this yet.

[thakis]

Thanks!

Common/Strings.cpp

41 ↗

(On Diff #165672)

__imp_ is also used for C names. But as far as I can tell, for C names a prefix of __imp_ doesn't necessarily mean that this is an import thunk for whatever follows it: the thunk for __declspec(dllimport) fand the name of a regular function void _imp__f() both get the name __imp__f: https://godbolt.org/z/sbObpp So while this is likely the right thing to do in C, it would only be a heuristic there.

If you want, I can add this for C too -- we might want to also don't strip the leading _that the compiler adds for C functions and chop off the @correction suffix that's added for some calling conventions if we wan to "demangle" C names. Currently we only demangle C++ names, so I only added this for C++ symbols.

[ruiu]

LGTM. I'm convinced that it makes sense to demangle only C++ names.

Common/Strings.cpp

43 ↗

(On Diff #165672)

Maybe this comment is too detailed for a first-time reader of this code? I'd just say that we demangle only C++ names.

[thakis]

Thanks! Changed the comment; landing.

Common/Strings.cpp

41 ↗

(On Diff #165672)

Somewhat entertainingly, the following program links and runs fine (with cl and link; 32-bit, for 64-bit it needs to be called __imp_f since no _ are inserted by the compiler there):

#include <stdio.h>

// Note: Declared only, not defined.
__declspec(dllimport) int f();

void g(void) { printf("hi\n"); }
void (*_imp__f)(void) = &g;

int main() { f(); }

lld/trunk/Common/Strings.cpp

Optional<std::string> lld::demangleItanium(StringRef Name) {
   if (!Buf)
     return None;
   std::string S(Buf);
   free(Buf);
   return S;
 }
 
 Optional<std::string> lld::demangleMSVC(StringRef Name) {
+  std::string Prefix;
+  if (Name.consume_front("__imp_"))
+    Prefix = "__declspec(dllimport) ";
+
+  // Demangle only C++ names.
   if (!Name.startswith("?"))
     return None;
+
   char *Buf = microsoftDemangle(Name.str().c_str(), nullptr, nullptr, nullptr);
   if (!Buf)
     return None;
   std::string S(Buf);
   free(Buf);
-  return S;
+  return Prefix + S;
 }
 
 StringMatcher::StringMatcher(ArrayRef<StringRef> Pat) {
   for (StringRef S : Pat) {
     Expected<GlobPattern> Pat = GlobPattern::create(S);
     if (!Pat)
       error(toString(Pat.takeError()));
     else

lld/trunk/test/COFF/undefined-symbol.s

 # REQUIRES: x86
 # RUN: llvm-mc -triple=x86_64-windows-msvc -filetype=obj -o %t.obj %s
 # RUN: not lld-link /out:%t.exe %t.obj 2>&1 | FileCheck %s
 
 # CHECK: error: undefined symbol: "int __cdecl foo(void)" (?foo@@YAHXZ)
 # CHECK-NEXT: >>> referenced by {{.*}}.obj:(main)
 # CHECK-NEXT: >>> referenced by {{.*}}.obj:(main)
 # CHECK-EMPTY:
 # CHECK-NEXT: error: undefined symbol: "int __cdecl bar(void)" (?bar@@YAHXZ)
 # CHECK-NEXT: >>> referenced by {{.*}}.obj:(main)
 # CHECK-NEXT: >>> referenced by {{.*}}.obj:(f1)
 # CHECK-EMPTY:
-# CHECK-NEXT: error: undefined symbol: "int __cdecl baz(void)" (?baz@@YAHXZ)
+# CHECK-NEXT: error: undefined symbol: "__declspec(dllimport) int __cdecl baz(void)" (__imp_?baz@@YAHXZ)
 # CHECK-NEXT: >>> referenced by {{.*}}.obj:(f2)
 
         .section        .text,"xr",one_only,main
 .globl main
 main:
   call  "?foo@@YAHXZ"
   call  "?foo@@YAHXZ"
   call  "?bar@@YAHXZ"
 
 f1:
   call  "?bar@@YAHXZ"
 .Lfunc_end1:
 
         .section        .text,"xr",one_only,f2
 .globl f2
 f2:
-  call  "?baz@@YAHXZ"
+  callq  *"__imp_?baz@@YAHXZ"(%rip)