IEEE support issues

IEEE support issues

[Alexander Aganichev]

Hi!

1. This code in ieee.c was fixed in CVS but then reverted:

==
  info->filename = filename;
  modname = strrchr (filename, '/');
  /* We could have a mixed forward/back slash case.  */
  backslash = strrchr (modname, '\\');
  if (backslash > modname)
    modname = backslash;

  if (modname != NULL)
    ++modname;
==

Could anyone please to replace at least `modname' with `filename' in the
second strrchr so binutils will not crash on IEEE producing? I really don't
care about signed pointers but this change is essential for me since we
regularly produce IEEE object files from COFF in our company ;-)

2. The second issue I found is that GCC and Binutils differently parses
size in .stab debug information: GCC produces sizes in bits and Binutils
assumes bytes. I don't know who is right (though it seems like GCC) but
this issue should be resolved. Currently I've added division by 8 to the
parse_stab_type() (stabs.c) as follows:
==
       switch (*attr)
         {
         case 's':
           size = atoi (attr + 1) / 8; /* adjust the size */
           if (size <= 0)
          size = -1;
           break;

         case 'S':
           stringp = true;
           break;

         default:
           /* Ignore unrecognized type attributes, so future
           compilers can invent new ones.  */
           break;
         }
==

--
Alexander Aganichev
Hypercom Europe Limited, Inc.
Software Engineer

Re: IEEE support issues

[Alan Modra]

On Sat, 1 Jul 2000, Alexander Aganichev wrote:

> 1. This code in ieee.c was fixed in CVS but then reverted:

No, your patch of 2000-06-20 wasn't reverted.  It was the incorrect
2000-06-25 patch to ar.c and bucomm.c that I reverted.

> 2. The second issue I found is that GCC and Binutils differently parses
> size in .stab debug information: GCC produces sizes in bits and Binutils
> assumes bytes. I don't know who is right (though it seems like GCC) but
> this issue should be resolved. Currently I've added division by 8 to the
> parse_stab_type() (stabs.c) as follows:

Thanks for debugging this.  Your patch looks OK to me, but in future
please provide patches using "diff -up" along with a plain text changelog
entry like the example below.  Changing binutils internal debug structures
to store size in bits is probably more correct, but a much larger change.
I'll check in the following:


binutils/ChangeLog
2000-07-01  Alexander Aganichev <AAganichev@hypercom.com>

	* stabs.c (parse_stab_type): Divide size in bits by 8 as binutils
	struct debug_type stores size in bytes.

Index: stabs.c
===================================================================
RCS file: /cvs/src/src/binutils/stabs.c,v
retrieving revision 1.4
diff -u -p -r1.4 stabs.c
--- stabs.c	2000/01/14 23:10:21	1.4
+++ stabs.c	2000/07/01 09:38:33
@@ -1261,6 +1261,7 @@ parse_stab_type (dhandle, info, typename
 	    {
 	    case 's':
 	      size = atoi (attr + 1);
+	      size /= 8;  /* Size is in bits.  We store it in bytes.  */
 	      if (size <= 0)
 		size = -1;
 	      break;

-- 
Linuxcare.  Support for the Revolution.

Where should gcc-avr info files go?

[Russ.Shaw]

Hi all,

After installing gcc-avr in mandrake-7.0, i have found some info files
that have avr help. However, i can't get them to appear in the main
info system. I'm not sure if the installation is supposed to update
the main info system, or whether you're meant to run info in some
different directory. Is there any kind of convention for this?


-- 
*******************************************
*   Russell Shaw, B.Eng, M.Eng(Research)  *
*      email: russell@webaxs.net          *
*      Australia                          *
*******************************************