This is the mail archive of the binutils@sourceware.org mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: 64-bit host failures, tic4x


On Mon, Feb 27, 2012 at 05:23:21PM +1030, Alan Modra wrote:
> +	  /* FIXME: subseg_text_p tests SEC_CODE which isn't in allowed
> +	     section flags.  See bfd/coff-tic4x.c target vecs.  */

Here's the obvious fix, but since I'm not a COFF expert I'm asking
for an OK to commit.  My admittedly superficial testing didn't show
any differences in section types.

bfd/
	* coff-tic4x.c (tic4x_coff0_vec, tic4x_coff0_beh_vec,
	tic4x_coff1_vec, tic4x_coff1_beh_vec, tic4x_coff2_vec,
	tic4x_coff2_beh_vec): Allow SEC_CODE and SEC_READONLY in
	section flags.
gas/
	* config/tc-tic4x.c: Remove alignment TODO comments.
	(tic4x_do_align): Enable subseg_text_p test.

Index: bfd/coff-tic4x.c
===================================================================
RCS file: /cvs/src/src/bfd/coff-tic4x.c,v
retrieving revision 1.16
diff -u -p -r1.16 coff-tic4x.c
--- bfd/coff-tic4x.c	18 Jul 2008 11:30:22 -0000	1.16
+++ bfd/coff-tic4x.c	27 Feb 2012 07:18:39 -0000
@@ -1,6 +1,6 @@
 /* BFD back-end for TMS320C4X coff binaries.
    Copyright 1996, 1997, 1998, 1999, 2000, 2002, 2003, 2005, 2007,
-   2008  Free Software Foundation, Inc.
+   2008, 2012 Free Software Foundation, Inc.
 
    Contributed by Michael Hayes (m.hayes@elec.canterbury.ac.nz)
 
@@ -274,19 +274,31 @@ tic4x_reloc_processing (relent, reloc, s
 
 
 /* TI COFF v0, DOS tools (little-endian headers).  */
-CREATE_LITTLE_COFF_TARGET_VEC(tic4x_coff0_vec, "coff0-tic4x", HAS_LOAD_PAGE, 0, '_', NULL, (PTR)&ticoff0_swap_table);
+CREATE_LITTLE_COFF_TARGET_VEC(tic4x_coff0_vec, "coff0-tic4x",
+			      HAS_LOAD_PAGE, SEC_CODE | SEC_READONLY, '_',
+			      NULL, &ticoff0_swap_table);
 
 /* TI COFF v0, SPARC tools (big-endian headers).  */
-CREATE_BIGHDR_COFF_TARGET_VEC(tic4x_coff0_beh_vec, "coff0-beh-tic4x", HAS_LOAD_PAGE, 0, '_', &tic4x_coff0_vec, (PTR)&ticoff0_swap_table);
+CREATE_BIGHDR_COFF_TARGET_VEC(tic4x_coff0_beh_vec, "coff0-beh-tic4x",
+			      HAS_LOAD_PAGE, SEC_CODE | SEC_READONLY, '_',
+			      &tic4x_coff0_vec, &ticoff0_swap_table);
 
 /* TI COFF v1, DOS tools (little-endian headers).  */
-CREATE_LITTLE_COFF_TARGET_VEC(tic4x_coff1_vec, "coff1-tic4x", HAS_LOAD_PAGE, 0, '_', &tic4x_coff0_beh_vec, (PTR)&ticoff1_swap_table);
+CREATE_LITTLE_COFF_TARGET_VEC(tic4x_coff1_vec, "coff1-tic4x",
+			      HAS_LOAD_PAGE, SEC_CODE | SEC_READONLY, '_',
+			      &tic4x_coff0_beh_vec, &ticoff1_swap_table);
 
 /* TI COFF v1, SPARC tools (big-endian headers).  */
-CREATE_BIGHDR_COFF_TARGET_VEC(tic4x_coff1_beh_vec, "coff1-beh-tic4x", HAS_LOAD_PAGE, 0, '_', &tic4x_coff1_vec, (PTR)&ticoff1_swap_table);
+CREATE_BIGHDR_COFF_TARGET_VEC(tic4x_coff1_beh_vec, "coff1-beh-tic4x",
+			      HAS_LOAD_PAGE, SEC_CODE | SEC_READONLY, '_',
+			      &tic4x_coff1_vec, &ticoff1_swap_table);
 
 /* TI COFF v2, TI DOS tools output (little-endian headers).  */
-CREATE_LITTLE_COFF_TARGET_VEC(tic4x_coff2_vec, "coff2-tic4x", HAS_LOAD_PAGE, 0, '_', &tic4x_coff1_beh_vec, COFF_SWAP_TABLE);
+CREATE_LITTLE_COFF_TARGET_VEC(tic4x_coff2_vec, "coff2-tic4x",
+			      HAS_LOAD_PAGE, SEC_CODE | SEC_READONLY, '_',
+			      &tic4x_coff1_beh_vec, COFF_SWAP_TABLE);
 
 /* TI COFF v2, TI SPARC tools output (big-endian headers).  */
-CREATE_BIGHDR_COFF_TARGET_VEC(tic4x_coff2_beh_vec, "coff2-beh-tic4x", HAS_LOAD_PAGE, 0, '_', &tic4x_coff2_vec, COFF_SWAP_TABLE);
+CREATE_BIGHDR_COFF_TARGET_VEC(tic4x_coff2_beh_vec, "coff2-beh-tic4x",
+			      HAS_LOAD_PAGE, SEC_CODE | SEC_READONLY, '_',
+			      &tic4x_coff2_vec, COFF_SWAP_TABLE);
Index: gas/config/tc-tic4x.c
===================================================================
RCS file: /cvs/src/src/gas/config/tc-tic4x.c,v
retrieving revision 1.33
diff -u -p -r1.33 tc-tic4x.c
--- gas/config/tc-tic4x.c	27 Feb 2012 06:52:20 -0000	1.33
+++ gas/config/tc-tic4x.c	27 Feb 2012 07:18:40 -0000
@@ -27,11 +27,6 @@
   o .align cannot handle fill-data-width larger than 0xFF/8-bits. It
     should be possible to define a 32-bits pattern.
 
-  o .align fills all section with NOP's when used regardless if has
-    been used in .text or .data. (However the .align is primarily
-    intended used in .text sections. If you require something else,
-    use .align <size>,0x00)
-
   o .align: Implement a 'bu' insn if the number of nop's exceeds 4
     within the align frag. if(fragsize>4words) insert bu fragend+1
     first.
@@ -2967,9 +2962,7 @@ tic4x_do_align (int alignment,
     {
       if (fill == NULL)
         {
-	  /* FIXME: subseg_text_p tests SEC_CODE which isn't in allowed
-	     section flags.  See bfd/coff-tic4x.c target vecs.  */
-          if (1 || subseg_text_p (now_seg))
+          if (subseg_text_p (now_seg))
 	    {
 	      char nop[4];
 

-- 
Alan Modra
Australia Development Lab, IBM


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]