> >> >  gdb_test "break mixed.adb:${bp_location} if light = green" \
> >> > -         "Breakpoint \[0-9\]* at .*: file .*/mixed.adb, line \[0-9\]*\\."
> >> > +         "Breakpoint \[0-9\]* at .*: file (.*/)?mixed.adb, line \[0-9\]*\\."
> > Isn't that the same as just:
> > 
> >  -         "Breakpoint \[0-9\]* at .*: file .*/mixed.adb, line \[0-9\]*\\."
> >  +         "Breakpoint \[0-9\]* at .*: file .*mixed.adb, line \[0-9\]*\\."
> > 
> > ?
> > 
> 
> Huh, sorry, brain malfunction.  My version would give out a spurious
> PASS with foomixed.adb, yours wouldn't.

That's the idea, yes. And to be more complete, you would get a full path
in the case where you built GDB out-of-tree, which you would get
a simple filename if when you build GDB in-tree.

> >> >  gdb_test "continue" \
> >> > -    "Breakpoint .*, pck\\.call_me \\(w=(w@entry=)?50\\) at .*/pck.adb:.*" \
> >> > +    "Breakpoint .*, pck\\.call_me \\(w=(w@entry=)?50\\) at (.*)?/pck.adb:.*" \
> >> >      "continue to call_me"
> > Likewise, I think this is a no-op.  Did you mean to put the / inside the
> > parens like in the other case?  If so I'd suggest:
> 
> This one still looks odd to me, though.

... and it is wrong. I don't understand how it could have slipped,
since I am supposed to have tested it in both in-tree and out-of-tree,
and this should have caused a failure in the in-tree case. Thanks
for spotting it!

I fixed this, removed the FIXME, and retested both in-tree and
out-of-tree, getting clean and identical results in both cases.

gdb/testsuite:

        * lib/ada.exp (target_compile_ada_from_dir): New function.
        (gdb_compile_ada): Reimplement avoiding the use of project files.
        * gdb.ada/gnat_ada.gpr: Delete.
        * gdb.ada/cond_lang.exp: Adjust test to make path before
        filename optional.
        * gdb.ada/small_reg_param.exp: Likewise.

Pushed to master.

Thanks again for the help, Pedro, and Merry Christmas :)
-- 
Joel