Re: Remote set thread breakpoint

[Andrew Cagney <cagney at gnu dot org>]

> Hi
>
> I'm thinking about extending the remote protocol for thread breakpoints.
> The easiest solution would be to just add another data field to the Z
> commands, like:
>
> Ztype,addr,length,threadid
>
> If threadid is given, gdbserver can use it. If it's -1 or not present (as in
> the present implementation) it's a global breakpoint.
>
> But if this would cause incompatibilities with existing parsers it may
> be better to create a new command, even if it's just Z5 and the rest
> is the same.
>
> What would be better? Or is something like that already in the pipe?

Yes [zZ]5 would be safer.  Would also be a good oportunity to formalize 
how to probe support for this packet - Z5?  Given things like vCont and 
its vCont? query, a [better?] alternative might be be vBP...

Just note that there is a small challenge here.  GDB internally assumes 
that breakpoints are global (it's a limitation / bug) - you'll need to 
also investigate  what needs to be changed closer to GDB's core.

Andrew