Re: [PATCH 1/5] Extended-remote exec events

[Don Breazeal <donb at codesourcery dot com>]

On 7/16/2015 9:35 AM, Yao Qi wrote:
> On 16/07/15 16:51, Don Breazeal wrote:
>> You make a good point, GDBserver doesn't handle that case.  I assume
>> that's what this is about:
>> ---
>> warning: Selected architecture i386 is not compatible with reported
>> target architecture i386:x86-64
>> ---
>> I'll investigate.
> 
> This messages shows that GDB (rather than GDBserver) doesn't handle
> that case.  AFAIK, GDBserver doesn't handle that case either.
> 
> I am working on patches create target description at the right time
> in GDBserver, derived from this patch
> https://sourceware.org/ml/gdb-patches/2015-07/msg00403.html
> Current GDBserver creates target description too early, if we use
> --wrapper option, GDBserver created target description according
> to wrapper program, instead of the program we want to debug, which
> is wrong.
> 

There is a difference between the native and gdbserver behavior with
multi-arch exec.  Native seems to handle multi-arch exec events:
-------------------------------------------------------------------
Reading symbols from ./execler64...done.
(gdb) b main
Breakpoint 1 at 0x4006a4: file execler.c, line 19.
(gdb) r
Starting program: /home/dbreazea/junk/execler64

Breakpoint 1, main (argc=1, argv=0x7fffffffe848) at execler.c:19
19	  printf ("starting %s\n", argv[0]);
(gdb) info reg
rax            0x7ffff7dd9ea8	140737351884456
rbx            0x0	0
rcx            0x0	0
---etc---
(gdb) catch exec
Catchpoint 2 (exec)
(gdb) c
Continuing.
starting /home/dbreazea/junk/execler64
in execler
process 5588 is executing new program: /home/dbreazea/junk/execee32
warning: the debug information found in "/lib/ld-2.11.1.so" does not
match "/lib/ld-linux.so.2" (CRC mismatch).


Catchpoint 2 (exec'd /home/dbreazea/junk/execee32), 0xf7fe0850 in ?? ()
   from /lib/ld-linux.so.2
(gdb) info reg
eax            0x0	0
ecx            0x0	0
---etc---
---------------------------------------------------------------------

While the gdbserver case looks like this, unfortunately:

---------------------------------------------------------------------
Reading symbols from ./execler64...done.
(gdb) tar ext localhost:51111
Remote debugging using localhost:51111
Reading symbols from target:/lib64/ld-linux-x86-64.so.2...(no debugging
symbols found)...done.
0x00007ffff7dddaf0 in ?? () from target:/lib64/ld-linux-x86-64.so.2
(gdb) b main
Breakpoint 1 at 0x4006a4: file execler.c, line 19.
(gdb) c
Continuing.

Breakpoint 1, main (argc=1, argv=0x7fffffffe9d8) at execler.c:19
19	  printf ("starting %s\n", argv[0]);
(gdb) info reg
rax            0x7ffff7dd9ea8	140737351884456
rbx            0x0	0
rcx            0x0	0
---etc---
(gdb) catch exec
Catchpoint 2 (exec)
(gdb) c
Continuing.
Thread 5561.5561 is executing new program: /home/dbreazea/junk/execee32
warning: Selected architecture i386 is not compatible with reported
target architecture i386:x86-64
warning: the debug information found in "target:/lib/ld-2.11.1.so" does
not match "target:/lib/ld-linux.so.2" (CRC mismatch).

Remote 'g' packet reply is too long: 000000000000000000000000000000...
---several 'g' packet errors
(gdb) q
A debugging session is active.

	Inferior 1 [process 5561] will be killed.

Quit anyway? (y or n) y
warning: Selected architecture i386 is not compatible with reported
target architecture i386:x86-64
--------------------------------------------------------------------

At first glance it looks like in linux_low_filter_event, the execing
inferior needs to be marked as a 'new_inferior'
(proc->priv->new_inferior) in order to do the right thing and call
the_low_target.arch_setup ().  That may require some re-ordering of
things in linux-low.c:linux_low_filter_event, since handle_extended_wait
and the exec event handling happen after the arch setup.

Do you think the work you are doing will address this, or should I
continue looking at a fix for the problem above?  They seem like they
are related, but separate issues.

thanks
--Don