[lldb-dev] Question about qC gdb request.

[Jean-Daniel Dupas]

Le 2 déc. 2011 à 01:47, Greg Clayton a écrit :

> 
> On Dec 1, 2011, at 2:55 PM, Jean-Daniel Dupas wrote:
> 
>> 
>> Le 1 déc. 2011 à 19:52, Greg Clayton a écrit :
>> 
>>> It was this way after misinterpreting what the GPLv2 version of GDB does (our darwin version) with the response. So it has been wrong for a while and does need to be fixed.
> 
> Part of the reason we did this too was the GDB remote specification docs:
> 
> 
> current thread	qC	Return the current thread id.
> reply QCpid	Where pid is a HEX encoded 16 bit process id.
> reply *	Any other reply implies the old pid.
> 
> It says "process id" is returned, not thread id…

OK. Look like it has been corrected though. 

>>> We need to add a new packet that gets information about the process (something like "qProcessInfo") that is being debugging in the case where we attach by name and need to know the process ID of what we attached to.
>> 
>> Isn't what qGetPid already does ? 
> 
> Not sure. Is there a home for the one true GDB remote specification? I don't see this packet in this:
> 
> http://sourceware.org/gdb/onlinedocs/gdb/General-Query-Packets.html#General-Query-Packets
> http://sources.redhat.com/gdb/current/onlinedocs/gdb.html#General-Query-Packets
> 
> Where are you seeing this packet? We don't look at any GPLv3 GDB sources at Apple, so we can't look there.

My bad. You're right, it's not in any specification. I just found it declared and supported in debugserver, but didn't checked if it was a lldb defined packet or a gdb one.

> Also, we want more than just the PID, we want architecture, pointer size, and vendor, OS, and other key value pairs, and thus we will soon implement the qProcessInfo packet.


Make sense.

-- Jean-Daniel