This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH 3/8] Set general_thread after restart
- From: Pedro Alves <palves at redhat dot com>
- To: Yao Qi <qiyaoltc at gmail dot com>
- Cc: gdb-patches at sourceware dot org
- Date: Fri, 24 Jul 2015 10:53:33 +0100
- Subject: Re: [PATCH 3/8] Set general_thread after restart
- Authentication-results: sourceware.org; auth=none
- References: <1437392126-29503-1-git-send-email-yao dot qi at linaro dot org> <1437392126-29503-4-git-send-email-yao dot qi at linaro dot org> <55B1718B dot 10100 at redhat dot com> <86fv4dhpc2 dot fsf at gmail dot com>
On 07/24/2015 10:33 AM, Yao Qi wrote:
>>> +# Test running programs using extended-remote.
>>
>> Comment looks stale. Looks like I missed pointing out the same
>> in patch #2.
>>
>
> I'll remove it.
I find these descriptions useful. Could you instead write something
like:
"Test restarting programs with the R packet."
?
>
>> Otherwise looks good to me.
>>
>> (I think it's likely we have lots of stale-data bugs on the
>> gdb side after R, as we don't resync much. It previously crossed my mind
>> that immediately after sending R, gdb should refresh all its
>> remote state anew, like if it had just disconnected and then reconnected.
>> That is, do most of what remote_start_remote does, except the
>> connection-specific details (qSupported, etc.)
>> Hard to justify the effort though -- I don't think I ever worked with
>> a stub that relies on R.)
>
> Even GDB refreshes all its state after sending R packet, we still need
> some way to test GDB and GDBserver with R packet used. Otherwise, it
> will be bit-rotten in the future.
Sounds like we're talking past each other.
Not sure what I said that made it sounds like that
idea would obviate the need for the test -- I think your new
test is great.
I meant something like gdb itself, around extended_remote_restart, calling
into a new function factored out from remote_start_remote.
This is because the R packet is documented as having no reply, like
'k', no doubt because it assumes the remote target can really hard reset
after the R packet. But let's forget it; hardly worth it to spend time
on it right now.
>
> GDBserver has already had an option --disable-packet, so that we can
> extend it to force GDBserver/GDB use R packet. However, I don't think
> we use --disable-packet much in our testing, at least I don't. Probably
> we can hack native-gdbserver.exp to run tests in a loop and pass
> different --disable-packet=FOO to GDBserver in each iteration.
Thanks,
Pedro Alves