Re: [patch] Suggest newer gdbserver if it has no qXfer:exec-file:read

[Jan Kratochvil <jan.kratochvil@redhat.com>]

On Tue, 22 Mar 2016 13:24:03 +0100, Pedro Alves wrote:
> On 03/19/2016 08:18 PM, Jan Kratochvil wrote:
> >    if (packet_support (PACKET_qXfer_exec_file) != PACKET_ENABLE)
> >-    return NULL;
> >+    {
> >+      warning (_("No executable has been specified (see the \"file\" command) "
> >+                 "and remote gdbserver does not "
> >+		 "support packet \"qXfer:exec-file:read\""
> >+		 " - please use FSF gdbserver version 7.10 or later."));
> >+      return NULL;
> >+    }
> 
> I think this will print the warning after connecting to any
> random stub, not just gdbserver.  Won't it be confusing
> to suggest FSF gdbserver in that case?

(1) I think this message can only appear during a mistake.  Is it right?
In fact this is my primary concern with this patch.
In such case I find any info better than no info.

(2) Still it may suggest they could for example implement qXfer:exec-file:read
in their gdbserver stub if appropriate.  I believe that people who use custom
gdbserver stub are more aware of how to fix it than normal
(=desktop/enterprise) OS developers who just try to debug some programs.

(3) Do you have a better idea?  One could add "if approproate" in that
message but I find that excessive.  One could detect FSF gdbserver
(if possible, I do not think it is, BTW it could be good to identify
variant+version of gdbserver over the protocol) but then still if it either is
or is not a FSF gdbserver that message may be relevant in some cases.


Thanks,
Jan