[RFC 0/2] Let's discuss moving gdbserver to top-level

[Simon Marchi]

On 2019-05-30 5:30 p.m., Tom Tromey wrote:
> I've wanted to move gdbserver to the top-level for a while now.  I
> think it will provide a decent benefit, mainly by letting gdb and
> gdbserver share their libiberty, gnulib, and "common" libraries --
> shaving off some build time.  It also will have the nice side effect
> of simplifying gdbserver's Makefile, fixing some existing bugs.
> 
> I have written the patches to do this, but before reworking them into
> submittable form, I thought I'd send a couple of initial patches and
> start a discussion.
> 
> This short series removes a couple of barriers to turning "common"
> into a library.  These are pretty much the only changes that could be
> sent in isolation.
> 
> The rest of the series is:
> 
> * Move gnulib to the top-level
> * Rename "common" to "gdbsupport" in preparation for a move
> * Move "gdbsupport" to top-level
> * Move gdbserver to top-level
> * Make gdbserver share the other top-level libraries
> 
> Most of the patches are pretty mechanical.  There are one or two
> hacks, primarily because this series doesn't move the "nat" directory.
> 
> Let me know what you think.
> 
> Tom

Just some questions about how things will work after such a move (at least, how it
is in your branch).

I suppose that there will be a top-level --enable-gdbserver/--disable-gdbserver.
switch like there is for other components?

If so, what happens if you do something like

  ./configure --host=x86 --target=arm --enable-gdb --enable-gdbserver

?  Currently, if the host and target architectures are different, gdbserver won't be built
(this check is done in gdb/configure.ac:2181).

Is gdbserver silently skipped, or does it error out saying that you can't enable gdbserver
if host != target?

Do you know about anything else in the binutils-gdb tree that is built to run on the target?

Simon