[PATCH 0/2] x86: don't recognize quoted symbol names as registers or operators

[PATCH 0/2] x86: don't recognize quoted symbol names as registers or operators

[Jan Beulich]

The first patch in particular is aiming at providing the foundation
to address gcc's bug target/53929.

1: don't recognize quoted symbol names as registers or operators
2: Intel: address quoted-symbol related FIXMEs

Jan

[PATCH 1/2] x86: don't recognize quoted symbol names as registers or operators

[Jan Beulich]

The concept of quoted symbols names was introduced pretty late. Utilize
it to allow access to symbols with names matching that of a register (or,
in Intel syntax, also an identifier-like operator).

This is primarily to aid gcc when generating Intel syntax output; see
their bug target/53929.
---
If desired, the i386_parse_name() could of course be restricted to
Intel syntax only. But 

--- a/gas/config/tc-i386.c
+++ b/gas/config/tc-i386.c
@@ -13868,6 +13868,13 @@ i386_parse_name (char *name, expressionS
   const reg_entry *r = NULL;
   char *end = input_line_pointer;
 
+  /* We only know the terminating character here.  It being double quote could
+     be the closing one of a quoted symbol name, or an opening one from a
+     following string (or another quoted symbol name).  Since the latter can't
+     be valid syntax for anything, bailing in either case is good enough.  */
+  if (*nextcharP == '"')
+    return 0;
+
   *end = *nextcharP;
   if (*name == REGISTER_PREFIX || allow_naked_reg)
     r = parse_real_register (name, &input_line_pointer);
--- a/gas/config/tc-i386-intel.c
+++ b/gas/config/tc-i386-intel.c
@@ -169,6 +169,10 @@ operatorT i386_operator (const char *nam
       return O_illegal;
     }
 
+  /* See the quotation related comment in i386_parse_name().  */
+  if (*pc == '"')
+    return O_absent;
+
   for (j = 0; i386_operators[j].name; ++j)
     if (strcasecmp (i386_operators[j].name, name) == 0)
       {
--- a/gas/testsuite/gas/i386/i386.exp
+++ b/gas/testsuite/gas/i386/i386.exp
@@ -105,6 +105,7 @@ if [gas_32_check] then {
     run_list_test "equ-bad"
     run_dump_test "divide"
     run_dump_test "quoted"
+    run_dump_test "quoted2"
     run_dump_test "unary"
     run_dump_test "padlock"
     run_dump_test "crx"
--- a/gas/testsuite/gas/i386/quoted.d
+++ b/gas/testsuite/gas/i386/quoted.d
@@ -18,4 +18,11 @@ Disassembly of section .text:
 [ 	]*[a-f0-9]+:[ 	]*ff 15 00 00 00 00[	 ]+call   \*0x0	[a-f0-9]+: (R_386_|dir)?32	x\(y\)
 [ 	]*[a-f0-9]+:[ 	]*26 ff 15 00 00 00 00[	 ]+call   \*%es:0x0	[a-f0-9]+: (R_386_|dir)?32	x\(y\)
 [ 	]*[a-f0-9]+:[ 	]*26 ff 15 00 00 00 00[	 ]+call   \*%es:0x0	[a-f0-9]+: (R_386_|dir)?32	x\(y\)
+[ 	]*[a-f0-9]+:[ 	]*b8 00 00 00 00[	 ]+mov    \$0x0,%eax	[a-f0-9]+: (R_386_|dir)?32	%eax
+[ 	]*[a-f0-9]+:[ 	]*a1 00 00 00 00[	 ]+mov    0x0,%eax	[a-f0-9]+: (R_386_|dir)?32	%eax
+[ 	]*[a-f0-9]+:[ 	]*a1 00 00 00 00[	 ]+mov    0x0,%eax	[a-f0-9]+: (R_386_|dir)?32	ecx
+[ 	]*[a-f0-9]+:[ 	]*a1 00 00 00 00[	 ]+mov    0x0,%eax	[a-f0-9]+: (R_386_|dir)?32	xmm0
+[ 	]*[a-f0-9]+:[ 	]*a1 00 00 00 00[	 ]+mov    0x0,%eax	[a-f0-9]+: (R_386_|dir)?32	not
+[ 	]*[a-f0-9]+:[ 	]*a1 00 00 00 00[	 ]+mov    0x0,%eax	[a-f0-9]+: (R_386_|dir)?32	and
+[ 	]*[a-f0-9]+:[ 	]*b8 00 00 00 00[	 ]+mov    \$0x0,%eax	[a-f0-9]+: (R_386_|dir)?32	edx
 #pass
--- a/gas/testsuite/gas/i386/quoted.s
+++ b/gas/testsuite/gas/i386/quoted.s
@@ -14,3 +14,13 @@ quoted:
 	call	*"x(y)"
 	call	*%es:"x(y)"
 	call	%es:*"x(y)"
+
+	mov	$"%eax", %eax
+	mov	"%eax", %eax
+
+	.intel_syntax noprefix
+	mov	eax, "ecx"
+	mov	eax, "xmm0"
+	mov	eax, "not"
+	mov	eax, "and"
+	mov	eax, offset "edx"
--- /dev/null
+++ b/gas/testsuite/gas/i386/quoted2.d
@@ -0,0 +1,15 @@
+#objdump: -r
+#name: i386 quoted symbols (data)
+# Mach-O relocations appear in inverse order
+#notarget: *-*-darwin
+
+.*: +file format .*
+
+RELOCATION RECORDS FOR \[\.data\]:
+OFFSET +TYPE +VALUE
+0+00 (R_386_|dir)?32 +%ebx
+0+04 (R_386_|dir)?32 +%rdx
+0+08 (R_386_|dir)?32 +eax
+0+0c (R_386_|dir)?32 +cr0
+0+10 (R_386_|dir)?32 +k0
+#pass
--- /dev/null
+++ b/gas/testsuite/gas/i386/quoted2.s
@@ -0,0 +1,9 @@
+	.data
+quoted:
+	.long	"%ebx"
+	.long	"%rdx"
+
+	.intel_syntax noprefix
+	.long	"eax"
+	.long	"cr0"
+	.long	"k0"

[PATCH 2/2] x86/Intel: address quoted-symbol related FIXMEs

[Jan Beulich]

If in a "word ptr <address>" or alike construct the "ptr" part is
double-quoted, it shouldn't be recognized as the specific keyword we're
looking for (just like we don't recognize double-quoted operator or
register names anymore). Be careful though to tell closing from opening
double-quotes, as a quoted symbol may follow right afterwards.

--- a/gas/config/tc-i386-intel.c
+++ b/gas/config/tc-i386-intel.c
@@ -188,15 +188,12 @@ operatorT i386_operator (const char *nam
 
   if (i386_types[j].name && *pc == ' ')
     {
+      const char *start = ++input_line_pointer;
       char *pname;
-      char c;
+      char c = get_symbol_name (&pname);
 
-      ++input_line_pointer;
-      c = get_symbol_name (&pname);
-
-      if (strcasecmp (pname, "ptr") == 0)
+      if (strcasecmp (pname, "ptr") == 0 && (c != '"' || pname == start))
 	{
-	  /* FIXME: What if c == '"' ?  */
 	  pname[-1] = *pc;
 	  *pc = c;
 	  if (intel_syntax > 0 || operands != 1)
@@ -204,9 +201,8 @@ operatorT i386_operator (const char *nam
 	  return i386_types[j].op;
 	}
 
-      if (strcasecmp (pname, "bcst") == 0)
+      if (strcasecmp (pname, "bcst") == 0 && (c != '"' || pname == start))
 	{
-	  /* FIXME: Again, what if c == '"' ?  */
 	  pname[-1] = *pc;
 	  *pc = c;
 	  if (intel_syntax > 0 || operands != 1