Wrong handling of -m32/-m64 [patch attached].

[markus-oberhumer-forks]

Hi,

first many thanks for your work on ldc - both LLVM and D are just great!

I'm completely new to ldc, but I needed the following patch against
ldc-0.15.1-src.tar.gz to make -m32/-m64 actually work.

cheers,
markus

Index: b/driver/ldmd.cpp
===================================================================
--- a/driver/ldmd.cpp
+++ b/driver/ldmd.cpp
@@ -945,7 +945,7 @@
     else if (p.debugInfo == Debug::pretendC) r.push_back("-gc");
     if (p.alwaysStackFrame) r.push_back("-disable-fp-elim");
     if (p.targetModel == Model::m32) r.push_back("-m32");
-    else if (p.targetModel == Model::m32) r.push_back("-m64");
+    else if (p.targetModel == Model::m64) r.push_back("-m64");
     if (p.profile) warning("CPU profile generation not yet supported by LDC.");
     if (p.verbose) r.push_back("-v");
     if (p.logTlsUse) warning("-vtls not yet supported by LDC.");
Index: b/driver/main.cpp
===================================================================
--- a/driver/main.cpp
+++ b/driver/main.cpp
@@ -992,6 +992,7 @@
         {
             error(Loc(), "cannot use both -m32 and -m64 options");
         }
+        bitness = ExplicitBitness::M64;
     }

     if (global.errors)

[dnadlinger]

Hi! Thanks a lot for the patch; I fixed the issue on master.

It really astonishes me how many bugs I managed to let slip into the LDMD code back in 2012 – must be the best example that late night coding isn't worth it in the long run.