Re: [RFA] add test for memattr, use get_number_or_range for memattr commands

[Michael Snyder <msnyder at vmware dot com>]

Joel Brobecker wrote:
> > 2011-02-21  Michael Snyder  <msnyder@vmware.com>
> >
> > 	* memattr.c (mem_enable_command): Use get_number_or_range.
> > 	(mem_disable_command): Ditto.
> > 	(mem_delete_command): Ditto.
> > 	(_initialize_mem): Tweak usage message to reflect multiple
> > 	arguments.
> >
> > 2011-02-21  Michael Snyder  <msnyder@vmware.com>
> >
> > 	* gdb.base/memattr.exp: New test.
> > 	* gdb.base/memattr.c: Test load for memattr.exp.
>
> Overall, the patch looks OK to me.
>
> > +gdb_test_multiple "info address mem1" "get address of mem1" {
> > +    -re "Symbol \"mem1\" is static storage at address ($hex).*$gdb_prompt $" {
> > +	set mem1start $expect_out(1,string)
> > +    }
> > +}
> [...]
> > +gdb_test_no_output "mem $mem1start $mem1end wo" "create mem region 1"
>
> I still think that, if the first quoted test does not pass, the second
> will cause the testcase to crash because mem1start is going to be
> undefined. This is what you're going to see when that happens:
>
>     ERROR: tcl error sourcing /[...]/memattr.exp
>     ERROR: can't read "mem1start": no such variable
>         while executing
>     "gdb_test_no_output "mem $mem1start $mem1end wo" "create mem region 1""
>
> But I'm OK with that, if that's OK with the others.

Hmmm, how about if I initialize mem1start etc. to -1?
Then the rest of the tests will fail but not crash.
Is that better?