Home Original page

usage of disassemble_info variables

Doug Evans dje@transmeta.com
Thu Dec 6 14:13:00 GMT 2001 
Ian Lance Taylor writes:
 > Doug Evans <dje@transmeta.com> writes:
 > 
 > > There seems to be an unwritten rule that a disassemble_info
 > > variable can't be a (non-static) local variable.
 > > Either that or INIT_DISASSEMBLE_mumble must initialize disassembler_options
 > > or each caller to print_insn_i386* is responsible for setting it.
 > > 
 > > We have various entry points into the x86 disassembler
 > > and I'm seeing a segv'ing because disassembler_options has garbage.
 > > 
 > > objdump initializes it.
 > > gdb doesn't but gets away with it because tm_print_insn_info is a global.
 > > 
 > > What's up?
 > 
 > I think that INIT_DISASSEMBLE_* should always be called on an instance
 > of disassembler_options before it is used.  From dis-asm.h:
 > 
 > /* Macro to initialize a disassemble_info struct.  This should be called
 >    by all applications creating such a struct.  */
 > 
 > I think this applies whether the disassembler_options is global,
 > static, or local.  If gdb doesn't initialize it, it should.

Heh.

Look at it this way.

Working code stopped working when i386-dis.c started refering
to disassemble_info.disassembler_options.

Is the bug that INIT_DISASSEMBLE_mumble wasn't changed to initialize
disassembler_options when disassembler_options was added
to disassemble_info?

Or must all callers of print_insn_i386* now start initializing
disassembler_options first (hint: they _already_ call INIT_DISASSEMBLE_INFO).

And for bonus points, will all such callers go through this again and again
and again each time similar things are added to dissemble_info?

["Boo! The fans cry foul as he pulls out the sledgehammer."]
More information about the Binutils mailing list