[PATCH] Fix bfd ABI string for MIPS

[PATCH] Fix bfd ABI string for MIPS

[Thiemo Seufer]

Hi All,

currently bfd returns for the MIPS ABI string "N32" for every ELF
object with header flag EF_MIPS_ABI2 set and "64" for every ELF
object which is 64bit.

This means E_MIPS_ABI_O64 and E_MIPS_ABI_EABI64 are named wrongly
as "64", and an object with EF_MIPS_ABI2 and arbitrary other flags
set is also wrongly called "N32".

The patch below fixes this.


Thiemo


2001-08-16  Thiemo Seufer <seufer@csv.ica.uni-stuttgart.de>

	/bfd/ChangeLog
	* elf32-mips.c (elf_mips_abi_name): Return the right ABI string for
	E_MIPS_ABI_O64 and E_MIPS_ABI_EABI64


diff -BurPX /bigdisk/dl/src/binutils-exclude src-orig/bfd/elf32-mips.c src/bfd/elf32-mips.c
--- src-orig/bfd/elf32-mips.c	Mon Jun 18 21:48:24 2001
+++ src/bfd/elf32-mips.c	Thu Jun 21 07:45:18 2001
@@ -1928,16 +1928,16 @@
 {
   flagword flags;
 
-  if (ABI_N32_P (abfd))
-    return "N32";
-  else if (ABI_64_P (abfd))
-    return "64";
-
   flags = elf_elfheader (abfd)->e_flags;
   switch (flags & EF_MIPS_ABI)
     {
     case 0:
-      return "none";
+      if (ABI_N32_P (abfd))
+	return "N32";
+      else if (ABI_64_P (abfd))
+	return "64";
+      else
+	return "none";
     case E_MIPS_ABI_O32:
       return "O32";
     case E_MIPS_ABI_O64:

Re: [PATCH] Fix bfd ABI string for MIPS

[Eric Christopher]

On 16 Aug 2001 03:41:48 +0200, Thiemo Seufer wrote:
> Hi All,
> 
> currently bfd returns for the MIPS ABI string "N32" for every ELF
> object with header flag EF_MIPS_ABI2 set and "64" for every ELF
> object which is 64bit.
> 
> This means E_MIPS_ABI_O64 and E_MIPS_ABI_EABI64 are named wrongly
> as "64", and an object with EF_MIPS_ABI2 and arbitrary other flags
> set is also wrongly called "N32".
> 
> The patch below fixes this.
> 
> 

I think this may have side effects.  How did you test this? gcc and gdb?

-eric


-- 
Look out behind you!

Re: [PATCH] Fix bfd ABI string for MIPS

[Andrew Cagney]

> On 16 Aug 2001 03:41:48 +0200, Thiemo Seufer wrote:
> 
>> Hi All,
>> 
>> currently bfd returns for the MIPS ABI string "N32" for every ELF
>> object with header flag EF_MIPS_ABI2 set and "64" for every ELF
>> object which is 64bit.
>> 
>> This means E_MIPS_ABI_O64 and E_MIPS_ABI_EABI64 are named wrongly
>> as "64", and an object with EF_MIPS_ABI2 and arbitrary other flags
>> set is also wrongly called "N32".
>> 
>> The patch below fixes this.
>> 
>> 
> 
> 
> I think this may have side effects.  How did you test this? gcc and gdb?
> 
> -eric
> 


Yes, I wondered the same thing.  Here, I think it doesn't have an effect 
  on GDB. GDB grubs aound in the bits in the elf header when determining 
the ABI.  However, caution is definitly advised.

	Andrew

Re: [PATCH] Fix bfd ABI string for MIPS

[Thiemo Seufer]

Andrew Cagney wrote:
> > On 16 Aug 2001 03:41:48 +0200, Thiemo Seufer wrote:
> > 
> >> Hi All,
> >> 
> >> currently bfd returns for the MIPS ABI string "N32" for every ELF
> >> object with header flag EF_MIPS_ABI2 set and "64" for every ELF
> >> object which is 64bit.

First of all, sorry for this bad description. It's not that bfd
returns this string to an client, it's use is internal to
elf32-mips.c. Having a closer look, it's in fact only used for
an error message about modules with incompatible ABI.

> >> This means E_MIPS_ABI_O64 and E_MIPS_ABI_EABI64 are named wrongly
> >> as "64", and an object with EF_MIPS_ABI2 and arbitrary other flags
> >> set is also wrongly called "N32".
> >> 
> >> The patch below fixes this.
> > 
> > I think this may have side effects.  How did you test this? gcc and gdb?
> > 
> > -eric
> > 
> 
> Yes, I wondered the same thing.  Here, I think it doesn't have an effect 
>   on GDB. GDB grubs aound in the bits in the elf header when determining 
> the ABI.  However, caution is definitly advised.

Sorry again. It definitely has no side effect except of putting out
the correct ABI names in the error message.


Thiemo

Re: [PATCH] Fix bfd ABI string for MIPS

[Thiemo Seufer]

Thiemo Seufer wrote:
[snip]
> > > I think this may have side effects.  How did you test this? gcc and gdb?
> > > 
> > > -eric
> > > 
> > 
> > Yes, I wondered the same thing.  Here, I think it doesn't have an effect 
> >   on GDB. GDB grubs aound in the bits in the elf header when determining 
> > the ABI.  However, caution is definitly advised.
> 
> Sorry again. It definitely has no side effect except of putting out
> the correct ABI names in the error message.

I'll check this in if there are no objections.


Thiemo

Re: [PATCH] Fix bfd ABI string for MIPS

[Eric Christopher]

> I'll check this in if there are no objections.
> 

No objections here. I was worried about Andrew more than anyone else :)

-eric

-- 
Look out behind you!

Re: [PATCH] Fix bfd ABI string for MIPS

[Andrew Cagney]

> 
> No objections here. I was worried about Andrew more than anyone else  [:)] 
> 

Which andrew?  If its me, it shouldn't affect GDB.  GDB doesn't use 
those strings.

	Andrew

Re: [PATCH] Fix bfd ABI string for MIPS

[Eric Christopher]

On 24 Aug 2001 16:39:51 -0400, Andrew Cagney wrote:
> > 
> > No objections here. I was worried about Andrew more than anyone else  [:)] 
> > 
> 
> Which andrew?  If its me, it shouldn't affect GDB.  GDB doesn't use 
> those strings.

Yup. You :) Glad to know that gdb doesn't use that stuff.

-eric

-- 
Look out behind you!