Re: [PATCH 1/2] GDB test suite: Add helper for locating core files

[Pedro Alves <palves@redhat.com>]

On 10/12/2017 06:00 PM, Andreas Arnez wrote:
> On Thu, Oct 12 2017, Pedro Alves wrote:
> 
>> On 10/09/2017 07:46 PM, Andreas Arnez wrote:
>>
>>> @@ -5883,7 +5883,7 @@ proc run_and_get_core {binfile {arg ""}} {
>>>  # specified.  Return that path name, or "" if no core file was found.
>>>  
>>>  proc find_core {binfile coredir {destcore ""}} {
>>> -    if {[is_remote target]} {
>>> +    if {![isnative]} {
>>>  	warning "Can not access remote core file."
>>>  	return ""
>>>      }
>>
>> This seems incorrect to me.  "isnative" only checks
>> if the build and target _triplets_ are the same.  So
>> foo-linux-gnu gdb x foo-linux-gnu gdbserver on separate
>> machine still returns isnative==true.
> 
> Exactly, that's why I originally wrote is_remote instead.  And I also
> wondered why the core dump tests check isnative.  Does anyone run the
> testsuite on a native remote setup?

I think most if not all isnative checks are stale/bogus,
and likely were introduced at a time when cross debugging,
including cross-core debugging wasn't that well supported.
I'd be delighted to see them cleaned up and replaced with
a more appropriate check.

> Right, this seems all a bit mixed up to me, similar to the confusing
> remote_file and remote_exec operations in core_find.

I wonder whether it'd be equally simple to move in the opposite
direction of making these operations use the right FOO
"remote_exec FOO", making it easier to actually support remote
core dump testing.  (Maybe not by default, since uploading the
cores may be slow, but with a board option/setting.)

Thanks,
Pedro Alves