cancel: [patch] Fix gdb.cp/gdb2495.exp regression with gcc-4.7 #4 [Re: [revert] Regression on PowerPC]

[Jan Kratochvil <jan dot kratochvil at redhat dot com>]

On Fri, 09 Mar 2012 00:23:45 +0100, Jan Kratochvil wrote:
> 2012-03-09  Jan Kratochvil  <jan.kratochvil@redhat.com>
> 
> 	* amd64-linux-tdep.c: Include inferior.h.
> 	(amd64_linux_init_abi): Set ON_STACK and i386_linux_push_dummy_code.
> 	* i386-linux-tdep.c (i386_linux_push_dummy_code): New function.
> 	(i386_linux_init_abi): Set ON_STACK and i386_linux_push_dummy_code.
> 	* i386-tdep.h (i386_linux_push_dummy_code): New declaration.

FYI cancelling this patch, it has some issues:

(1) As i386-tdep.c already has OS-independent i386_push_dummy_call I believe
    push_dummy_code can be also put into i386-tdep.c.

(2) At least on RHEL-5 i386 there is a regression:
    FAIL: gdb.base/call-signal-resume.exp: continue to program exit

#0  null_hand_call () at ./gdb.base/call-signals.c:56
#1  <function called from gdb>
#2  0xf7fe0430 in __kernel_vsyscall ()
#3  0xf7e7d236 in kill () from /lib/libc.so.6
#4  0x080484f4 in gen_signal () at ./gdb.base/call-signals.c:35
#5  0x08048574 in main () at ./gdb.base/call-signals.c:81
->
#0  null_hand_call () at ./gdb.base/call-signals.c:56
#1  <function called from gdb>
#2  0xf7fe0430 in __kernel_vsyscall ()
#3  0xf7e7d236 in kill () from /lib/libc.so.6
#4  0xf70484f4 in ?? ()
#5  0x00004b03 in ?? ()
#6  0x00000006 in ?? ()
#7  0xffffd068 in ?? ()
#8  0x08048574 in main () at ./gdb.base/call-signals.c:81

This is because (a) the dummy frame has no associated DWARF/EH frame unwind
info and i386_push_dummy_call does not setup proper frame pointer in the new
frame.  Moreover I believe it should be setup with 0 frame pointer / 0 return
address to indicate end of unwinding for possible inferior unwinders, IMO
inferior should not unwind through the dummy frame.  And then therefore GDB
needs new frame_unwind to properly unwind that 0 return address / frame
pointer.


Regards,
Jan