Skip to contentSkip to conversation

GCC release 8 support for gcc-plugins

PATCH In review [email protected] 3 participants 4 messages Feb 4, 2018 – Feb 13, 2018 Gcc plugins

lore
Share & tools
lore
#0 lore raw

PATCH 2/2 GCC release 8 support for gcc-plugins

For reasons totally beyond my understanding, gcc 8 changed the order of two structure member, which leads to an error:

HOSTCXX -fPIC scripts/gcc-plugins/latent_entropy_plugin.o
scripts/gcc-plugins/latent_entropy_plugin.c:269:1: error: designator order for field 'attribute_spec::affects_type_identity' does not match declaration order in 'attribute_spec'
};
^
make[1]: *** [scripts/Makefile.host:141: scripts/gcc-plugins/latent_entropy_plugin.o] Error 1

Maybe we'll get lucky and they'll undo that change before official release, so all this ad-crock #ifdef-ery won't be needed. Do we know any gcc experts to ask?

--- diff --git a/scripts/gcc-plugins/latent_entropy_plugin.c b/scripts/gcc-plugins/latent_entropy_plugin.c index 65264960910d..15a6df3592c5 100644

scripts/gcc-plugins/latent_entropy_plugin.c +9 -2
@@ -262,10 +262,17 @@ static struct attribute_spec latent_entropy_attr = {
 	.decl_required			= true,
 	.type_required			= false,
 	.function_type_required		= false,
+#if BUILDING_GCC_VERSION >= 8000
+	.affects_type_identity		= false,
 	.handler			= handle_latent_entropy_attribute,
+#else
 #if BUILDING_GCC_VERSION >= 4007
-	.affects_type_identity		= false
-#endif
+	.handler			= handle_latent_entropy_attribute,
+	.affects_type_identity		= false,
+#else
+	.handler			= handle_latent_entropy_attribute,
+#endif /* >= 4007 */
+#endif /* >= 8000 */
 };
 
 static void register_attributes(void *event_data __unused, void *data __unused)
scripts/gcc-plugins/structleak_plugin.c +10 -3
index 3f8dd4868178..f700bc059050 100644
@@ -64,10 +64,17 @@ static struct attribute_spec user_attr = {
 	.decl_required		= false,
 	.type_required		= false,
 	.function_type_required	= false,
-	.handler		= handle_user_attribute,
+#if BUILDING_GCC_VERSION >= 8000
+	.affects_type_identity		= false,
+	.handler			= handle_latent_entropy_attribute,
+#else
 #if BUILDING_GCC_VERSION >= 4007
-	.affects_type_identity	= true
-#endif
+	.handler			= handle_latent_entropy_attribute,
+	.affects_type_identity		= false,
+#else
+	.handler			= handle_latent_entropy_attribute,
+#endif /* >= 4007 */
+#endif /* >= 8000 */
 };
 
 static void register_attributes(void *event_data, void *data)
scripts/gcc-plugins/randomize_layout_plugin.c +40 -12

80 lines not on this page · raw message

#1 lore raw
Quoted 7 lines · On Mon, Feb 5, 2018 at 4:34 AM, <[email protected]> wrote:

For reasons totally beyond my understanding, gcc 8 changed the order of two structure member, which leads to an error:

HOSTCXX -fPIC scripts/gcc-plugins/latent_entropy_plugin.o
scripts/gcc-plugins/latent_entropy_plugin.c:269:1: error: designator order for field 'attribute_spec::affects_type_identity' does not match declaration order in 'attribute_spec'
};
^

FWIW, my gcc 8 snapshot reports:

scripts/gcc-plugins/latent_entropy_plugin.c:269:1: sorry,
unimplemented: non-trivial designated initializers not supported
};
^

But yeah, it seems that initializer _order_ matters here. Uuuugh.

I think instead of static initializers, we can just move this into the register function, then order won't matter again. e.g.:

(quoted diff) +0 -0

b/scripts/gcc-plugins/latent_entropy_plugin.c
index 65264960910d..6836b8cd9fc4 100644

scripts/gcc-plugins/latent_entropy_plugin.c +0 -0
@@ -255,21 +255,21 @@ static tree handle_latent_entropy_attribute(tree

*node, tree name,

        return NULL_TREE;

}

-static struct attribute_spec latent_entropy_attr = {

-       .name                           = "latent_entropy",
-       .min_length                     = 0,
-       .max_length                     = 0,
-       .decl_required                  = true,
-       .type_required                  = false,
-       .function_type_required         = false,
-       .handler                        = handle_latent_entropy_attribute,

-#if BUILDING_GCC_VERSION >= 4007
- .affects_type_identity = false
-#endif
-};
+static struct attribute_spec latent_entropy_attr;

static void register_attributes(void *event_data __unused, void *data __unused) {

+       latent_entropy_attr.name                        = "latent_entropy";
+       latent_entropy_attr.min_length                  = 0;
+       latent_entropy_attr.max_length                  = 0;
+       latent_entropy_attr.decl_required               = true;
+       latent_entropy_attr.type_required               = false;
+       latent_entropy_attr.function_type_required      = false;
+       latent_entropy_attr.handler     = handle_latent_entropy_attribute;

+#if BUILDING_GCC_VERSION >= 4007
+ latent_entropy_attr.affects_type_identity = false;
+#endif
+

        register_attribute(&latent_entropy_attr);

}

(pardon whitespace damage...)

-Kees

Signature · 2 lines
Kees Cook
Pixel Security
#2 lore raw
In reply to [email protected]
Quoted 2 lines · On Sun, Feb 04, 2018 at 12:34:38PM -0500, [email protected] wrote:

For reasons totally beyond my understanding, gcc 8 changed the order of two structure member, which leads to an error:

I'm a little late on this, but isn't the whole point of C99 initializers that field order should not matter at all?

#3 lore raw
Quoted 6 lines · Christoph Hellwig
Quoted 2 lines · On Sun, Feb 04, 2018 at 12:34:38PM -0500, [email protected] wrote:

For reasons totally beyond my understanding, gcc 8 changed the order of two structure member, which leads to an error:

I'm a little late on this, but isn't the whole point of C99 initializers that field order should not matter at all?

For C99, this is true. The plugins are built differently and follow C++ conventions, which, for reasons I cannot understand, are ordered. :( Regardless, the fix actually results in fewer lines of code, so this is fine.

-Kees

Signature · 2 lines
Kees Cook
Pixel Security