build: fix spurious mksnapshot crashes for good

A variety of gcc bugs made mksnapshot crash with either a segmentation fault
or a 'pure virtual method callled' run-time error.

After much wailing and gnashing of teeth I managed to deduce that the bugs
show up when:

  1. gcc 4.5.2 for i386-pc-solaris2.11 is used and -fstrict-aliasing is
     enabled, or

  2. gcc version 4.4.6 for x86_64-redhat-linux is used and
     -ffunction-sections -finline-functions at -O2 or higher is enabled

Therefore, disable -ffunction-sections and -fdata-sections unconditionally
and disable -fstrict-aliasing only on Solaris.

The -ffunction-sections and -fdata-sections switches were nonsense anyway
because we don't link with -Wl,--gc-sections.

2 files changed, 8 insertions, 15 deletions

diff --git a/common.gypi b/common.gypi
index 8d6041418..e12469ff4 100644
--- a/common.gypi
+++ b/common.gypi
@@ -1,6 +1,5 @@
 {
   'variables': {
-    'node_no_strict_aliasing%': 0,   # turn off -fstrict-aliasing
     'visibility%': 'hidden',         # V8's visibility setting
     'target_arch%': 'ia32',          # set v8's target architecture
     'host_arch%': 'ia32',            # set v8's host architecture
@@ -42,7 +41,12 @@
         },
       },
       'Release': {
-        'cflags': [ '-O3', '-fdata-sections', '-ffunction-sections' ],
+        # Do *NOT* enable -ffunction-sections or -fdata-sections again.
+        # We don't link with -Wl,--gc-sections so they're effectively no-ops.
+        # Worse, they trigger very nasty bugs in some versions of gcc, notably
+        # v4.4.6 on x86_64-redhat-linux (i.e. RHEL and CentOS).
+        'cflags!': [ '-ffunction-sections', '-fdata-sections' ],
+        'cflags': [ '-O3' ],
         'conditions': [
           ['target_arch=="x64"', {
             'msvs_configuration_platform': 'x64',
@@ -52,9 +56,6 @@
             # pull in V8's postmortem metadata
             'ldflags': [ '-Wl,-z,allextract' ]
           }],
-          ['node_no_strict_aliasing==1', {
-            'cflags': [ '-fno-strict-aliasing' ],
-          }],
         ],
         'msvs_settings': {
           'VCCLCompilerTool': {
diff --git a/configure b/configure
index 5e60c55f4..b7d8498ab 100755
--- a/configure
+++ b/configure
@@ -288,13 +288,6 @@ def configure_node(o):
 
   cc_version, is_clang = compiler_version()
 
-  # turn off strict aliasing if gcc < 4.6.0 unless it's llvm-gcc
-  # see http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45883
-  # see http://code.google.com/p/v8/issues/detail?id=884
-  no_strict_aliasing = int(not(is_clang or cc_version >= (4,6,0)))
-  o['variables']['v8_no_strict_aliasing'] = no_strict_aliasing
-  o['variables']['node_no_strict_aliasing'] = no_strict_aliasing
-
   # clang has always supported -fvisibility=hidden, right?
   if not is_clang and cc_version < (4,0,0):
     o['variables']['visibility'] = ''
@@ -303,9 +296,8 @@ def configure_node(o):
   # systems, since it won't work.  (The MacOS build process is different than
   # SunOS, and we haven't implemented it.)
   if sys.platform.startswith('sunos'):
-    o['variables']['node_use_dtrace'] = b(not options.without_dtrace);
-    # Strict aliasing causes problems with the V8 snapshots on SunOS
-    o['variables']['strict_aliasing'] = b(False);
+    o['variables']['node_use_dtrace'] = b(not options.without_dtrace)
+    o['variables']['v8_no_strict_aliasing'] = 1 # work around compiler bug
   elif b(options.with_dtrace) == 'true':
     raise Exception('DTrace is currently only supported on SunOS systems.')
   else: