Closed froydnj closed 4 years ago
I skimmed the patch a bit and nothing jumped out at me. I think it seems reasonable.
Maybe that's overthinking it (probably is for gdb-minidump MVP), and the only reasonable host I/O implementation is one that essentially does host libc calls
I sometimes consider writing a federated gdbserver. In a case like this the I/O implementation would forward to some other server. So, I think this has to be abstracted somehow.
One idea might be to have a second Handler
-like thing and have process_packets_from
take another argument.
I went ahead and implemented a libc-handler in my fork
Merged a modified version of this on master, huge thanks!
I think I still need to write a few more tests, and github tells me I need to rebase, but this is at least at the point that things build and pass tests; feedback would be useful.
I don't know where to stick the handlers.
Handler
is the obvious place, but it feels like the host I/O stuff is conceptually separate, and people should be able to mix-and-match with handlers and not have to rewrite aHandler
implementation. Maybe that's overthinking it (probably is forgdb-minidump
MVP), and the only reasonable host I/O implementation is one that essentially does hostlibc
calls.