Joel, I'm a little vague about gdb_test_multiple, but is the attached 
better?

Michael

Joel Brobecker wrote:
>> Attached is a test case for it -- it will have a number of
>> XFAILS without this patch, which will become PASSES with the patch.
> 
> I looked at the testcase, and noticed a couple of things:
> 
>> #   Copyright 1995, 1996, 1997, 1999, 2007, 2008, 2009
> 
> I was surprised by the copyright years, but I then read the contents
> of the testcase that mentions alpha-osf3.  That's when I realized
> that this file is inspired by sigall.exp... I supposed that this
> was the right thing to do, although it probably does not matter
> because I believe that these headers are actually not legally
> significant.
> 
>> if [target_info exists gdb,nosignals] {
>>     verbose "Skipping sigall-reverse.exp because of nosignals."
>>     continue
>> }
> 
> I wonder why we do a continue here, whereas we do a return elsewhere:
> 
>> if ![target_info exists gdb,can_reverse] {
>>     return
>> }
> 
> I wish we had a cookbook for writing testcases, I always forget what
> we're supposed to do :-(. Anyone knows if this is significant?
> 
>> 	send_gdb "continue\n"
>> 	if { $thissig == "IO" } {
>> 	    setup_xfail "i*86-pc-linuxoldld-gnu" "i*86-pc-linuxaout-gnu"
>> 	}
>> 	gdb_expect {
> [...]
> 
> IMO, the send_gdb/gdb_expect sequences in this script should be converted
> to using test_gdb_multiple. I'd rather we avoid send_gdb/gdb_expect
> if we can.
>