a problem with ld options

[Nick Clifton]

Hi Chenyu,

>>  readelf try.o -r
>>  readelf try.out -r
>>  The first readelf works well, but the second exits with segment fault.
> 
> 
> Well, that's at least two bugs.  readelf shouldn't segfault, and ld -q
> ought to work.  :-( 

Attached is a patch for the first bug.  It fixes readelf so that it no 
longer assumes that the string table offset for dynamic relocs will be a 
valid offset into the dynamic string table.  Instead it checks and 
prints a suitable error message if a corrupt index is found.  I will be 
applying this patch shortly.

Cheers
   Nick

binutils/ChangeLog

2004-10-25  Nick Clifton  <nickc@redhat.com>

	* readelf.c (dynamic_strings_length): New global variable.
	(VALID_DYNAMIC_NAME, GET_DYNAMIC_NAME): New macros for accessing
	strings in the dynamic string table.
	(process_section_headers): Initialise dynamic_strings_length.
	(process_dynamic_section): Likewise.
	(process_object): Reset dynamic_string_length when the buffer is freed.
	(dynamic_sections_mips_val): Use the new macros.
	(process_dynamic_section): Likewise.
	(process_version_sections): Likewise.
	(process_symbol_table): Likewise.
	(process_syminfo): Likewise.
	(process_mips_specific): Likewise.
	(dump_relocations): Add a new parameter 'strtablen' and use this
	to verify that string offset in a given reloc is valid.  Print a
	suitable error message otherwise.
	(process_relocs): Pass the new argument to dump_relocations.

-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: readelf.c.patch
URL: <https://sourceware.org/pipermail/binutils/attachments/20041025/3685a258/attachment.ksh>