Re: [RFA] gdb.base, f*.exp to l*.exp, send_gdb vs. gdb_test

[Michael Snyder <msnyder@vmware.com>]

Pedro Alves wrote:
> On Friday 21 May 2010 21:15:56, Michael Snyder wrote:
>> +    set test "unpatch child, unpatched parent breakpoints from child"
>> +    gdb_test_multiple "continue" $test {
>> +       -re "at exit.*$gdb_prompt $" {
>> +           pass "$test"
>> +       }
>> +       -re "SIGTRAP.*$gdb_prompt $" {
>> +           fail "$test"
>> +
>> +           # Explicitly kill this child, so we can continue gracefully
>> +           # with further testing...
>> +           send_gdb "kill\n"
>> +           gdb_expect {
>> +               -re ".*Kill the program being debugged.*y or n. $" {
>> +                   send_gdb "y\n"
>> +                   gdb_expect -re "$gdb_prompt $" {}
>> +               }
>> +           }
>> +       }
>> +       -re ".*$gdb_prompt $" {
>> +           fail "$test (unknown output)"
>> +       }
>> +       timeout {
>> +           fail "$test (timeout)"
>> +       }
> 
> I guess these last two cases could be deleted too?  Doesn't
> matter much --- okay anyway.

Good catch.


>> --- long_long.exp       1 Jan 2010 07:32:01 -0000       1.30
>> +++ long_long.exp       21 May 2010 20:12:01 -0000
>> @@ -61,15 +61,9 @@ if { ![runto known_types] } then { 
> ...
>> -set target_bigendian_p 1
> 
> I think you should still set this, so if the test fails,
> the following tests referencing this don't error out reading
> a non-existing variable.  Wait, the gdb_test_bi function in
> this file, the only user of this variable, isn't itself used
> anywhere.  Huh?  This means the variable could be garbage
> collected instead, and this gdb_test_multiple below simplified.

Yeah, in fact we don't need it at all.

Thanks for the review.