Patch set updated for coreboot: xcompile: Tell gcc that we need to access NULL pointers occassionally

[Patrick Georgi]

Patrick Georgi ([e-mail]) just uploaded a new patch set to gerrit, which you can find at https://review.coreboot.org/12298

-gerrit

commit bd1edbcc7cf3e3e075ecdb6f34e6b47cfe40d174
Author: Patrick Georgi <[e-mail]>
Date:   Tue Nov 3 15:30:02 2015 +0100

    xcompile: Tell gcc that we need to access NULL pointers occassionally
    
    gcc developers consider NULL pointers to be so bad that they add a trap
    instruction (ud2 on x86, something similar on ARM) when gcc encounters
    one.
    
    We need them every now and then (eg. via/cx700 raminit), so tell the
    compiler not to treat us like a userspace program.
    
    Change-Id: I1a062bfa55548ed0c2d11196633c6f63e83b807c
    Signed-off-by: Patrick Georgi <[e-mail]>
---
 util/xcompile/xcompile | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/util/xcompile/xcompile b/util/xcompile/xcompile
index 76c3698..8718c67 100755
--- a/util/xcompile/xcompile
+++ b/util/xcompile/xcompile
@@ -161,6 +161,11 @@ detect_special_flags() {
 	testcc "$GCC"      "$CFLAGS_GCC -Wl,--build-id=none" &&
 		CFLAGS_GCC="$CFLAGS_GCC -Wl,--build-id=none"
 
+	# Without this flag, gcc's __builtin_trap() traps accesses to NULL
+	# that may well be intended in RAM init code.
+	testcc "$GCC"      "$CFLAGS_GCC -fno-delete-null-pointer-checks"&&
+		CFLAGS_GCC="$CFLAGS_GCC -fno-delete-null-pointer-checks"
+
 	case "$architecture" in
 	x86)
 		# Always build for i686 -- no sse/mmx instructions since SMM