Re: [PATCH] Improve MI inferior output check / mi-console.exp

[Pedro Alves <palves at redhat dot com>]

On 10/28/2013 02:19 AM, Luis Machado wrote:
> Hi,
> 
> The situation here is that the output from the target/inferior arrives
> differently (or not at all) depending on permutations of local/remote
> and the no-inferior-io setting.
> 
> The mi-console.c testcase writes a string 'Hello \"!' one char at a time.
> 
> In the case of a spawned inferior, to which we connect, we see that 
> string. In the case of remote/semi-hosted target we see the following:
> 
> @"H"
> @"e"
> @"l"
> @"l"
> @"o"
> @" "
> @"\\"
> @"\""
> @"!"
> @"\r"
> @"\n"
> 
> In the case of a remote target with noinferiorio set, we assume no 
> access to the output and thus this test is unsupported.
> 
> The attached patch modifies mi-console.exp to furnish the two different
> patterns, and mi-support.exp to check for more cases.
> 
> Regression tested on x86-64.
> 
> Ok?
> 
> 
> mi_output.diff
> 
> 
> 2013-10-27  Iain Sandoe  <iain@codesourcery.com>
> 	    Luis Machado  <lgustavo@codesourcery.com>
> 
>        gdb/testsuite
>        * lib/mi-support.exp (mi_gdb_test): Expect different formats
>        of inferior output for remote and native sessions.
>        * gdb.mi/mi-console.exp: Check for additional inferior output
>        pattern.
> 
> diff --git a/gdb/testsuite/gdb.mi/mi-console.exp b/gdb/testsuite/gdb.mi/mi-console.exp
> index b7643e7..d7b6db4 100644
> --- a/gdb/testsuite/gdb.mi/mi-console.exp
> +++ b/gdb/testsuite/gdb.mi/mi-console.exp
> @@ -37,10 +37,20 @@ if  { [gdb_compile "${srcdir}/${subdir}/${srcfile}" "${binfile}" executable {deb
>  mi_run_to_main
>  
>  # Next over the hello() call which will produce lots of output
> +# The output we get from the target depends on how it is hosted.
> +# If we are local, we just see the executable's output.
> +# If we are remote (and the configuration supports it), we see the
> +# target responses encapsulated by gdb's annotation.
> +if [target_info exists use_gdb_stub] {

Hmm, this predicate doesn't look right.

# This test only works when talking to a target that routes its output
# through GDB.  Check that we're either talking to a simulator or a
# remote target.

And indeed I think the sim will route output likewise:

 remote-sim.c:  ui_file_write (gdb_stdtarg, buf, len);
 remote-sim.c:  gdb_flush (gdb_stdtarg);
 remote-sim.c:      fputs_unfiltered (b, gdb_stdtargerr);
 remote-sim.c:  gdb_flush (gdb_stdtargerr);

Also, when testing in extended-remote mode, use_gdb_stub
will be off.

So it seems to me that e.g., with the sim, the test
will now timeout?

-- 
Pedro Alves