Re: [PATCH] Fix MI dprintf-insert not printing when a location is pending.

[Keith Seitz <keiths@redhat.com>]

On 03/27/2015 05:34 AM, Antoine Tremblay wrote:
> On 03/26/2015 03:10 PM, Keith Seitz wrote:
>> On 03/26/2015 09:47 AM, Antoine Tremblay wrote:
>>> +# Set pending dprintf via MI.
>>> +mi_gdb_test "-dprintf-insert -f pendfunc1 \"hello\"" \
>>> +
>>> ".*\\^done,bkpt={number=\"1\",type=\"dprintf\",disp=\"keep\",enabled=\"y\",addr=\"<PENDING>\",pending=\"pendfunc1\",times=\"0\",original-location=\"pendfunc1\"}"
>>>
>>> \
>>> +    "mi set dprintf"
>>> +
>>> +mi_gdb_test "-break-insert $bp_location1" ".*" "mi insert breakpoint
>>> bp_location1"
>>
>> Is it possible to use mi_make_breakpoint for these tests?
>
> Unfortunately for the dprintf one mi_make_breakpoint doesn't support
> pending breakpoints, it creates something like :
> bkpt={number="2",type=".*",disp=".*",enabled=".*",addr=".*",func=".*",
>         file=".*/myfile.c",fullname=".*",line="3",thread-groups=\[.*\],
>         times="0".*original-location=".*"}
>
> But with pending funcs it should be pending= ...

Ah, I see. Thank you for pointing that out. I *thought* that I did that 
already!

> It could be the subject of another patch to add that support.
>

Indeed.

> I used mi_create_breakpoint for the other breakpoint now
>

Excellent.

>>> +mi_run_cmd
>>> +
>>> +set msg "mi dprintf"
>>> +gdb_expect {
>>> +    -re ".*~\"hello\"" {
>>> +    pass $msg
>>> +    }
>>> +    -re ".*$mi_gdb_prompt$" {
>>> +    fail $msg
>>> +    }
>>> +    timeout {
>>> +    fail $msg
>>> +    }
>>> +}
>>
>> This a pretty common test suite idiom, I think. Can mi_gdb_test be used
>> instead of gdb_expect?
>
> That I can't since mi_gdb_test requires a command and in this case I'm
> just doing expect on that comes after mi_run_cmd, there's no command
> associated with it..

While not common in gdb.mi (with mi_gdb_test), the rest of gdb.* uses 
the idiom:

   gdb_test "" ...

all the time.

Looking at the description of mi_gdb_test:

# mi_gdb_test COMMAND PATTERN MESSAGE [IPATTERN] -- send a command to gdb;
#   test the result.
#
# COMMAND is the command to execute, send to GDB with send_gdb.  If
#   this is the null string no command is sent.

[Very quickly] Browsing through the code of that function, it seems like 
it should work. If it does not, would you please file a bug? It 
*definitely* should work.

Keith