Re: [PATCH 0/7] Do not require "file" commands for remote targets

[Pedro Alves <palves at redhat dot com>]

On 04/15/2015 01:02 PM, Gary Benson wrote:
> Pedro Alves wrote:
>> On 04/01/2015 12:22 PM, Gary Benson wrote:
>>> Patch 7 causes GDB to attempt to locate and open the executable
>>> in remote-target cases without multiprocess extensions:
>>
>> Should be "with multiprocess extensions".  Without those, we don't
>> know the target's pid, and thus can't use pid-to-exec-file.
> 
> How about:
> 
>   Patch 7 causes GDB to attempt to locate and open the executable
>   for remote targets not using extended-remote:

That's better.

Though it affects extended-remote just the same.  E.g.,:

  $ gdbserver --multi :9999 /bin/sh
  ...

  (gdb) target extended-remote :9999
  ...
  Reading symbols from target:/bin/bash...


Or without passing a binary to gdbserver:

  $ gdbserver --multi :9999
  ...

  (gdb) target extended-remote :9999
  ...
  (gdb) set remote exec-file /bin/sh
  ...
  (gdb) run
  ...
  (gdb) disconnect
  (gdb) quit

now reconnect in extended-remote with another gdb instance:

  $ gdb
  ...
  (gdb) target extended-remote :9999
   ...
  Reading symbols from target:/bin/bash...


So even more correct even would be to say that patch 7 causes GDB
to attempt to locate and open the executable of processes already
being debugged by the remote target, on initial connection,
with the simplest example being:

  bash$ gdb -q
  (gdb) target remote | gdbserver - /bin/sh
  Remote debugging using | gdbserver - /bin/sh
  Process /bin/sh created; pid = 32166
  stdin/stdout redirected
  Remote debugging using stdio
  Reading symbols from target:/bin/bash...

Thanks,
Pedro Alves