public inbox for binutils@sourceware.org
 help / color / mirror / Atom feed
* bogus "ignoring changed section type" warning
@ 2009-08-06 14:36 Jan Beulich
  2009-08-07  0:51 ` Alan Modra
  0 siblings, 1 reply; 5+ messages in thread
From: Jan Beulich @ 2009-08-06 14:36 UTC (permalink / raw)
  To: binutils

Isn't this warning rather odd for code like this:

	.section	.init.rodata,"a"
	.section	.init.rodata,"a",@progbits

especially since with just the first directive, the section ends up being
progbits anyway (due to elf_fake_sections() deriving the type from the
section flags)? Wouldn't it hence make sense to guess an unspecified
section type from the flags right away, or alternatively allow modifying
SHT_NULL (in obj_elf_change_section()) by a later, more explicit
directive?

Thanks, Jan

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: bogus "ignoring changed section type" warning
  2009-08-06 14:36 bogus "ignoring changed section type" warning Jan Beulich
@ 2009-08-07  0:51 ` Alan Modra
  2009-08-28 16:13   ` Jan Beulich
  0 siblings, 1 reply; 5+ messages in thread
From: Alan Modra @ 2009-08-07  0:51 UTC (permalink / raw)
  To: Jan Beulich; +Cc: binutils

On Thu, Aug 06, 2009 at 03:35:58PM +0100, Jan Beulich wrote:
> Isn't this warning rather odd for code like this:
> 
> 	.section	.init.rodata,"a"
> 	.section	.init.rodata,"a",@progbits
> 
> especially since with just the first directive, the section ends up being
> progbits anyway (due to elf_fake_sections() deriving the type from the
> section flags)?

Yes, there ought to be a warning on the first directive rather than on
the second one.

> Wouldn't it hence make sense to guess an unspecified
> section type from the flags right away,

Probably.

-- 
Alan Modra
Australia Development Lab, IBM

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: bogus "ignoring changed section type" warning
  2009-08-07  0:51 ` Alan Modra
@ 2009-08-28 16:13   ` Jan Beulich
  2009-08-30 18:32     ` Alan Modra
  0 siblings, 1 reply; 5+ messages in thread
From: Jan Beulich @ 2009-08-28 16:13 UTC (permalink / raw)
  To: Alan Modra; +Cc: binutils

>>> Alan Modra <amodra@bigpond.net.au> 07.08.09 02:51 >>>
>On Thu, Aug 06, 2009 at 03:35:58PM +0100, Jan Beulich wrote:
>> Isn't this warning rather odd for code like this:
>> 
>> 	.section	.init.rodata,"a"
>> 	.section	.init.rodata,"a",@progbits
>> 
>> especially since with just the first directive, the section ends up being
>> progbits anyway (due to elf_fake_sections() deriving the type from the
>> section flags)?
>
>Yes, there ought to be a warning on the first directive rather than on
>the second one.

I'm afraid this wouldn't be welcome to users: Just looking at the Linux
sources, there are quite a few uses without type, expecting it to default
to @progbits.

>> Wouldn't it hence make sense to guess an unspecified
>> section type from the flags right away,
>
>Probably.

Like this (2.19.1 based; I'd move it to mainline and formally submit if it seems
right)?

Jan

--- binutils-2.19.1/bfd/elf-bfd.h	2008-08-21 01:28:58.000000000 +0200
+++ 2.19.1/bfd/elf-bfd.h	2009-08-28 16:24:09.000000000 +0200
@@ -2128,6 +2128,8 @@ extern bfd_boolean _bfd_elf_map_sections
 
 extern bfd_boolean _bfd_elf_is_function_type (unsigned int);
 
+extern int bfd_elf_get_default_section_type (flagword);
+
 extern Elf_Internal_Phdr * _bfd_elf_find_segment_containing_section
   (bfd * abfd, asection * section);
 
--- binutils-2.19.1/bfd/elf.c	2008-12-23 14:54:48.000000000 +0100
+++ 2.19.1/bfd/elf.c	2009-08-28 16:28:43.000000000 +0200
@@ -2485,6 +2485,18 @@ _bfd_elf_init_reloc_shdr (bfd *abfd,
   return TRUE;
 }
 
+/* Return the default section type based on the passed in section flags.  */
+
+int
+bfd_elf_get_default_section_type (flagword flags)
+{
+  if ((flags & SEC_ALLOC) != 0
+      && ((flags & (SEC_LOAD | SEC_HAS_CONTENTS)) == 0
+	  || (flags & SEC_NEVER_LOAD) != 0))
+    return SHT_NOBITS;
+  return SHT_PROGBITS;
+}
+
 /* Set up an ELF internal section header for a section.  */
 
 static void
@@ -2534,12 +2546,8 @@ elf_fake_sections (bfd *abfd, asection *
      asect->flags.  */
   if ((asect->flags & SEC_GROUP) != 0)
     sh_type = SHT_GROUP;
-  else if ((asect->flags & SEC_ALLOC) != 0
-	   && (((asect->flags & (SEC_LOAD | SEC_HAS_CONTENTS)) == 0)
-	       || (asect->flags & SEC_NEVER_LOAD) != 0))
-    sh_type = SHT_NOBITS;
   else
-    sh_type = SHT_PROGBITS;
+    sh_type = bfd_elf_get_default_section_type (asect->flags);
 
   if (this_hdr->sh_type == SHT_NULL)
     this_hdr->sh_type = sh_type;
--- binutils-2.19.1/gas/config/obj-elf.c	2008-04-23 15:54:56.000000000 +0200
+++ 2.19.1/gas/config/obj-elf.c	2009-08-28 16:18:10.000000000 +0200
@@ -674,6 +685,8 @@ obj_elf_change_section (const char *name
     {
       symbolS *secsym;
 
+      if (type == SHT_NULL)
+	type = bfd_elf_get_default_section_type (flags);
       elf_section_type (sec) = type;
       elf_section_flags (sec) = attr;
 
--- binutils-2.19.1/gas/testsuite/gas/elf/section5.l	2004-12-21 14:09:55.000000000 +0100
+++ 2.19.1/gas/testsuite/gas/elf/section5.l	2009-08-28 16:37:15.000000000 +0200
@@ -2,7 +2,6 @@
 .*:7: Warning: .*
 .*:7: Warning: .*
 .*:10: Warning: .*
-.*:13: Warning: .*
 .*:16: Warning: .*
 .*:18: Warning: .*
 .*:20: Warning: .*


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: bogus "ignoring changed section type" warning
  2009-08-28 16:13   ` Jan Beulich
@ 2009-08-30 18:32     ` Alan Modra
  2009-08-31 14:14       ` Jan Beulich
  0 siblings, 1 reply; 5+ messages in thread
From: Alan Modra @ 2009-08-30 18:32 UTC (permalink / raw)
  To: Jan Beulich; +Cc: binutils

On Fri, Aug 28, 2009 at 04:06:47PM +0100, Jan Beulich wrote:
> Like this (2.19.1 based; I'd move it to mainline and formally submit if it seems
> right)?

Yes, preapproved.

-- 
Alan Modra
Australia Development Lab, IBM

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: bogus "ignoring changed section type" warning
  2009-08-30 18:32     ` Alan Modra
@ 2009-08-31 14:14       ` Jan Beulich
  0 siblings, 0 replies; 5+ messages in thread
From: Jan Beulich @ 2009-08-31 14:14 UTC (permalink / raw)
  To: Alan Modra; +Cc: binutils

>>> Alan Modra <amodra@bigpond.net.au> 30.08.09 07:54 >>>
>On Fri, Aug 28, 2009 at 04:06:47PM +0100, Jan Beulich wrote:
>> Like this (2.19.1 based; I'd move it to mainline and formally submit if it seems
>> right)?
>
>Yes, preapproved.

Just for the record, this is the patch I'm about to commit.

Jan

bfd/
2009-08-31  Jan Beulich  <jbeulich@novell.com>

	* elf-bfd.h (bfd_elf_get_default_section_type): Declare.
	* elf.c (bfd_elf_get_default_section_type): New.
	(elf_fake_sections): Use bfd_elf_get_default_section_type.

gas/
2009-08-31  Jan Beulich  <jbeulich@novell.com>

	* config/obj-elf.c (obj_elf_change_section): Set default type
	by calling bfd_elf_get_default_section_type.

gas/testsuite/
2009-08-31  Jan Beulich  <jbeulich@novell.com>

	* gas/elf/section5.l: Remove no longer issued warning pattern.

--- 2009-08-31/bfd/elf-bfd.h	2009-07-23 15:06:29.000000000 +0200
+++ 2009-08-31/bfd/elf-bfd.h	2009-08-31 10:07:16.000000000 +0200
@@ -2108,6 +2108,8 @@ extern bfd_boolean _bfd_elf_map_sections
 
 extern bfd_boolean _bfd_elf_is_function_type (unsigned int);
 
+extern int bfd_elf_get_default_section_type (flagword);
+
 extern Elf_Internal_Phdr * _bfd_elf_find_segment_containing_section
   (bfd * abfd, asection * section);
 
--- 2009-08-31/bfd/elf.c	2009-08-31 09:54:15.000000000 +0200
+++ 2009-08-31/bfd/elf.c	2009-08-31 10:07:16.000000000 +0200
@@ -2422,6 +2422,18 @@ _bfd_elf_init_reloc_shdr (bfd *abfd,
   return TRUE;
 }
 
+/* Return the default section type based on the passed in section flags.  */
+
+int
+bfd_elf_get_default_section_type (flagword flags)
+{
+  if ((flags & SEC_ALLOC) != 0
+      && ((flags & (SEC_LOAD | SEC_HAS_CONTENTS)) == 0
+	  || (flags & SEC_NEVER_LOAD) != 0))
+    return SHT_NOBITS;
+  return SHT_PROGBITS;
+}
+
 /* Set up an ELF internal section header for a section.  */
 
 static void
@@ -2471,12 +2483,8 @@ elf_fake_sections (bfd *abfd, asection *
      asect->flags.  */
   if ((asect->flags & SEC_GROUP) != 0)
     sh_type = SHT_GROUP;
-  else if ((asect->flags & SEC_ALLOC) != 0
-	   && (((asect->flags & (SEC_LOAD | SEC_HAS_CONTENTS)) == 0)
-	       || (asect->flags & SEC_NEVER_LOAD) != 0))
-    sh_type = SHT_NOBITS;
   else
-    sh_type = SHT_PROGBITS;
+    sh_type = bfd_elf_get_default_section_type (asect->flags);
 
   if (this_hdr->sh_type == SHT_NULL)
     this_hdr->sh_type = sh_type;
--- 2009-08-31/gas/config/obj-elf.c	2009-08-31 09:54:18.000000000 +0200
+++ 2009-08-31/gas/config/obj-elf.c	2009-08-31 10:07:16.000000000 +0200
@@ -685,6 +685,8 @@ obj_elf_change_section (const char *name
     {
       symbolS *secsym;
 
+      if (type == SHT_NULL)
+	type = bfd_elf_get_default_section_type (flags);
       elf_section_type (sec) = type;
       elf_section_flags (sec) = attr;
 
--- 2009-08-31/gas/testsuite/gas/elf/section5.l	2005-01-18 10:43:34.000000000 +0100
+++ 2009-08-31/gas/testsuite/gas/elf/section5.l	2009-08-31 10:07:16.000000000 +0200
@@ -2,7 +2,6 @@
 .*:7: Warning: .*
 .*:7: Warning: .*
 .*:10: Warning: .*
-.*:13: Warning: .*
 .*:16: Warning: .*
 .*:18: Warning: .*
 .*:20: Warning: .*



^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2009-08-31 12:01 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-08-06 14:36 bogus "ignoring changed section type" warning Jan Beulich
2009-08-07  0:51 ` Alan Modra
2009-08-28 16:13   ` Jan Beulich
2009-08-30 18:32     ` Alan Modra
2009-08-31 14:14       ` Jan Beulich

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).