Re: [PATCH] gdb/record-full: disable range stepping when resuming threads

[Simon Marchi <simon.marchi@efficios.com>]

On 5/2/23 11:57, Bruno Larsen wrote:
> On 02/05/2023 17:03, Tom Tromey wrote:
>>>>>>> "Bruno" == Bruno Larsen via Gdb-patches <gdb-patches@sourceware.org> writes:
>> Bruno> For future improvement, I'm wondering: is gdbserver aware that the
>> Bruno> execution is being recorded at all?
>>
>> I don't think so, or at least, not for full record.  (It seems to have
>> some btrace knowledge.)
>>
>> Bruno> That way it could record the
>> Bruno> execution by itself and once the stop is reported, it sends the
>> Bruno> recorded instructions (which I hope should be faster). I'm asking
>> Bruno> before adding it to my backlog, just in case it is very difficult to
>> Bruno> do.
>>
>> I guess at minimum it would require making it possible to build the
>> record machinery into gdbserver, and also adding the needed requests to
>> the remote protocol.
> 
> Ah, that's unfortunate. I was hoping it would already bt there and this would've mostly been a protocol addition. Oh well... thanks for taking the time to answer :-)

Also, I think that the built-in record mechanism (the record-full
target) is mostly obsolete at this point.  It hasn't been updated in
ages to support new CPU instructions, and doesn't support multi-threaded
programs.  To anybody using the record-full target, I would suggest
trying out something like rr [1] instead.

Simon

[1] https://rr-project.org/