[PATCH][RFC] Re: sparse handles int64_t type wrong

[Christopher Li <sparse@chrisli.org>]

I took a stab at it.  It seems that direct_declarator() does handle
attributes, but it does not change the ctype to the declarator.
It is kind of awkward that the type is already build when it hit
the attribute in the end. It need to go back to overwrite
the already parsed type.

I have this patch seems make the test case happy. I am not sure
this is correct or not.
 
Chris
 

Index: sparse/parse.c
===================================================================
--- sparse.orig/parse.c	2006-12-12 21:12:29.000000000 -0800
+++ sparse/parse.c	2006-12-12 21:12:47.000000000 -0800
@@ -837,6 +837,10 @@ static struct token *handle_attributes(s
 			struct ctype thistype = { 0, };
 			token = attribute_specifier(token->next, &thistype);
 			apply_ctype(token->pos, &thistype, ctype);
+			if (is_int_type(ctype->base_type) && ctype->modifiers & MOD_SPECIFIER) {
+				ctype->base_type = ctype_integer(ctype->modifiers);
+				ctype->modifiers &= ~MOD_SPECIFIER;
+			}
 			continue;
 		}
 		if (match_idents(token, &asm_ident, &__asm_ident, &__asm___ident)) {
Index: sparse/show-parse.c
===================================================================
--- sparse.orig/show-parse.c	2006-12-12 21:12:29.000000000 -0800
+++ sparse/show-parse.c	2006-12-12 21:12:47.000000000 -0800
@@ -98,7 +98,7 @@ const char *modifier_string(unsigned lon
 	const char *res,**ptr, *names[] = {
 		"auto", "register", "static", "extern",
 		"const", "volatile", "[signed]", "[unsigned]",
-		"[char]", "[short]", "[long]", "[long]",
+		"[char]", "[short]", "[long]", "[long long]",
 		"[typdef]", "[structof]", "[unionof]", "[enum]",
 		"[typeof]", "[attribute]", "inline", "[addressable]",
 		"[nocast]", "[noderef]", "[accessed]", "[toplevel]",
Index: sparse/test-parsing.c
===================================================================