I am using Eclipse CDT and the GNU ARM QEMU simulator. Eclipse starts it own gdb session with the gdbserver running on QEMU. I want to also be able to connect from an OSX terminal for more fine tuned control. I also want to be able to use the GDB/MI from a Node.js script to grab useful data like register status, etc. and visualise it in a webpage.
I can connect fine when running QEMU simulator standalone with my own gdb session, but I wanted to know if its possible to piggy back onto Eclipse's gdb, or make a second connection to gdbserver.
Thanks!
From https://sourceware.org/gdb/onlinedocs/gdb/Server.html (20.3.1.2):
When gdbserver stays running, gdb can connect to it again later. Such reconnecting is useful for features like disconnected tracing. For completeness, at most one gdb can be connected at a time.
So far I could not figure it out and it seems from the docs that it is not possible.
The solution I am using now is to run the QEMU simulator manually from the terminal. The GNU ARM Eclipse debugger allows specifiying a gdbserver host:port, which I can point at the standalone QEMU which works fine.
To allow simulatenous gdb sessions, I could create a custom TCP server for Eclipse to connect to instead of connecting directly to gdbserver. This TCP server would forward all traffic from Eclipse to the gdbserver, and I can also listen to and send messages to gdb over this channel.
The end result being that you can debug and set breakpoints in Eclipse as usual, but also interact with gdb any way you want to such as through a web interface, native OSX terminal, etc.
There is no built-in way to do this. There is a project to implement it, but I don't know how active it might be.
Your idea of having a kind of proxy might work. It's not really deeply different, I think, from the approach proposed for the multi-client project. If you're only interested in data that is already likely to go over the wire, another similar approach might be to insert a tee somewhere and just filter out the replies that are interesting.