llvm dev - May 2008 - [LLVMdev] [PATCH] fix for FreeBSD/powerpc build breakage

All,

The following patch fixes the build breakage on FreeBSD/powerpc:


Index: lib/Target/PowerPC/PPCJITInfo.cpp
==================================================================---
lib/Target/PowerPC/PPCJITInfo.cpp	(revision 51432)
+++ lib/Target/PowerPC/PPCJITInfo.cpp	(working copy)
@@ -72,7 +72,7 @@
  extern "C" void PPC64CompilationCallback();

  #if (defined(__POWERPC__) || defined (__ppc__) || defined(_POWER))  
&& \
-    !defined(__ppc64__)
+    !(defined(__ppc64__) || defined(__FreeBSD__))
  // CompilationCallback stub - We can't use a C function with inline  
assembly in
  // it, because we the prolog/epilog inserted by GCC won't work for  
us.  Instead,
  // write our own wrapper, which does things our way, so we have  
complete control


The patch is a bit of a kluge, but I don't know which of the
preprocessor defines are used on Darwin and/or Linux. FreeBSD
has all of the following:
	#define __ppc__ 1
	#define __PPC__ 1
	#define _ARCH_PPC 1
	#define __PowerPC__ 1
	#define __powerpc__ 1

A related define that could also be used to select the right
inline assembly statement is:
	#define __GNUCLIKE_ASM 3

In any case, register naming in inline assembly on FreeBSD is
the same as the Linux version, so maybe the condition can be
simplified and based on the OS?

FYI,

-- 
Marcel Moolenaar
xcllnt at mac.com

On May 22, 2008, at 11:47 AM, Marcel Moolenaar wrote:
> All,
>
> The following patch fixes the build breakage on FreeBSD/powerpc:
Thanks, applied!
http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20080519/062896.html
> The patch is a bit of a kluge, but I don't know which of the
> preprocessor defines are used on Darwin and/or Linux. FreeBSD
> has all of the following:
Yeah, that code is gross.  I'm not sure what it is trying to figure  
out.  If it just wants darwin/ppc for abi reasons, we could just test  
PPC & darwin.  I'm not sure exactly what it is trying to do though.

-Chris