> I'm thinking the right thing to do, here is to enhance select_event_lwp
> to look for threads that received a fork/vfork event first, and report
> that event to gdb ahead of all the other kinds of events.

That seems to work well. Attached is the patch doing this.

The commit's revision log gives I think a reasonable overview of
what the issue is and how the patch fixes it. But more information
that I found while I was investigating this which is only tangential
to the issue can be found in the following message, exchanged on
this list:

    https://www.sourceware.org/ml/gdb-patches/2016-05/msg00207.html
    https://www.sourceware.org/ml/gdb-patches/2016-05/msg00209.html
    https://www.sourceware.org/ml/gdb-patches/2016-06/msg00391.html

gdb/gdbserver/ChangeLog:

        * linux-low.c (select_fork_vfork_lwp_callback): New function.
        (select_event_lwp): Give priority fork/vfork events over all
        other types of events.

gdb/testsuite/ChangeLog:

        * gdb.ada/bp_and_fork: New test.

Tested on x86_64-linux.
OK to apply?

Thanks!
-- 
Joel