This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH, gdbserver] Pass tracepoint to 'add_traceframe_block'.
- From: Pedro Alves <palves at redhat dot com>
- To: Yao Qi <yao at codesourcery dot com>
- Cc: gdb-patches at sourceware dot org
- Date: Fri, 11 Jan 2013 18:50:12 +0000
- Subject: Re: [PATCH, gdbserver] Pass tracepoint to 'add_traceframe_block'.
- References: <1355971951-2598-1-git-send-email-yao@codesourcery.com> <50EB043D.600@redhat.com> <50EC307A.5000608@codesourcery.com>
On 01/08/2013 02:43 PM, Yao Qi wrote:
>> But what about:
>>
>>> >- # Tracepoint hit count is optional, so pass it either way.
>>> >-
>> ?
>>
>> And it seems like traceframe usage as just as optional as
>> hit count, both being reported by qTP, which the target
>> may not implement?
>
> I didn't fully understand "optional" in the line of comment.
The qTP packet was added after tracing support existed in the
RSP. It's not a mandatory packet. Stubs exist in the field
that don't support that packet. GDB should still work with
those.
>
> The per-tracepoint hit counts/traceframe usage and qTP was added by this patch,
>
> [PATCH v2] Tracing notes and metadata
> http://sourceware.org/ml/gdb-patches/2011-11/msg00484.html
>
> and the corresponding tests (for hit count) were added by this patch as well.
> IMO, the hit count and traceframe usage is *not* optional, because the same
> patch adds the support.
Not sure I understand.
> Do we have to worry about other stubs or old GDBserver that
> don't support qTP?
Yes, it's a good principle to follow.
--
Pedro Alves