Re: tdep/1089: gdb HC11/HC12 crashes after info frame command

[Andrew Cagney <ac131313 at redhat dot com>]

The following reply was made to PR tdep/1089; it has been noted by GNATS.

From: Andrew Cagney <ac131313 at redhat dot com>
To: Stephane Carrez <stcarrez at nerim dot fr>
Cc: Stephane dot Carrez at nerim dot fr, gdb-gnats at sources dot redhat dot com
Subject: Re: tdep/1089: gdb HC11/HC12 crashes after info frame command
Date: Wed, 26 Feb 2003 16:03:02 -0500

 Can you here do a:
 
 (gdb) set debug frame 1
 
 (I just added this command)
 
 > (gdb) b main
 > Breakpoint 1 at 0x4410: file /opt/gnu/gdb-main-integrate/gdb/testsuite/gdb.asm/asmsrc1.s, line 25.
 > (gdb) run
 > Starting program: /build/fsf-m68hc11/gdb-m6811-tui/gdb/testsuite/gdb.asm/asm-source
 > 
 > Breakpoint 1, main () at /opt/gnu/gdb-main-integrate/gdb/testsuite/gdb.asm/asmsrc1.s:29
 > 29              gdbasm_several_nops
 > Current language:  auto; currently asm
 
 You should here see the reason why the unwind failed.
 
 > (gdb) info frame
 > Stack frame at 0x0:
 >  pc = 0x4410 in main (/opt/gnu/gdb-main-integrate/gdb/testsuite/gdb.asm/asmsrc1.s:29);
 >     saved pc 0x4410
 >  source language asm.
 > 
 > Program received signal SIGSEGV, Segmentation fault.
 > 0x0811ded4 in m68hc11_frame_args_address (frame=0x82ae218)
 >     at /opt/gnu/gdb-main-integrate/gdb/m68hc11-tdep.c:413
 > 413       addr = get_frame_base (frame) + get_frame_extra_info (frame)->size + STACK_CORRECTION + 2;
 > (gdb) p *frame
 > $1 = {frame = 0, pc = 17424, level = -1, type = NORMAL_FRAME, saved_regs = 0x0, extra_info = 0x0,
 > 
 > *********** BUG: extra_info is null in frame
 > 
 >   context = 0x0, unwind_cache = 0x82ae25c, unwind = 0x82091e8, pc_unwind_cache_p = 1,
 >   pc_unwind_cache = 17424, id_unwind_cache_p = 0, id_unwind_cache = {base = 0, pc = 0},
 >   next = 0x82ae218, prev_p = 1, prev = 0x0} 
 
 Two problems:
 
 - For some reason GDB is failing to create the inner-most frame from the 
 sentinel frame - the above is an assembler file.  That leads to your 
 code being handed a sentinel frame (doing an init_extra_frame_info on 
 that doesn't make sense).  The funtction get_prev_frame() is doing this 
 and hopefully `set debug frame 1' will let us know where.
 
 - The info frame code should be better at handling sentinel frames.  But 
 lets not worry about that.
 
 Andrew