Re: [PATCH 4/5] gdb/testsuite: change newline patterns used in gdb_test

[Simon Marchi <simark@simark.ca>]

On 4/28/23 17:50, Andrew Burgess wrote:
> Simon Marchi <simark@simark.ca> writes:
> 
>> On 4/28/23 11:57, Simon Marchi via Gdb-patches wrote:
>>>> I've pushed the patch below which fixes all the regressions I see when
>>>> using the native-gdbserver board.
>>>>
>>>> I'm still testing with the native-extended-gdbserver board, but I'm a
>>>> little short of time, so might not be able to fix any regressions that
>>>> throws up until next week.  So I've pushed this fix for now, and will
>>>> follow up with any additional fixes I find later.
>>>>
>>>> Sorry for causing the regressions.
>>>
>>> Thanks, I'll let you know if there are more remaining failures.
>>>
>>> Simon
>>>
>>
>> I still see these two, which are likely related:
>>
>> FAIL: gdb.trace/collection.exp: collect register locals collectively: run trace experiment: stop trace experiment
> 
> I think we should accept this regression.  Here's the gdb.log output:
> 
>   (gdb) PASS: gdb.trace/collection.exp: collect register locals collectively: run trace experiment: run trace experiment
>   tstop
>   Trace is not running.
>   (gdb) FAIL: gdb.trace/collection.exp: collect register locals collectively: run trace experiment: stop trace experiment
> 
> I changed this test form this:
> 
> 	gdb_test "tstop" \
> 	    "\[\r\n\]+" \
> 	    "stop trace experiment"
> 
> to this:
> 
> 	gdb_test_no_output "tstop" \
> 	    "stop trace experiment"
> 
> The 'Trace is not running.' error is what causes the test to fail.  This
> test has a bunch of other FAILs, one of which is the 'tstart' preceeding
> the above 'tstop':
> 
>   (gdb) PASS: gdb.trace/collection.exp: collect register locals collectively: run trace experiment: advance to begin
>   tstart
>   locf has been optimized out of existence.
>   locd has been optimized out of existence.
>   Expression pieces exceed word size
>   (gdb) FAIL: gdb.trace/collection.exp: collect register locals collectively: run trace experiment: start trace experiment
> 
> So clearly, the test script is expecting the tracing to start, but it
> doesn't.  Catching things like the above -- where we were missing a
> warning/error from GDB was exactly the point of my patch -- so I think
> this additional failure is acceptable.

That makes sense (it would have been obvious if I had looked at it, but
didn't really have time).  I'll add it to my list of expected failures,
since it's just the consequence of an existing failure.

>> FAIL: gdb.trace/pending.exp: ftrace works: stop trace experiment
> 
> I'm not able to reproduce this one.  On my local machine the
> pending_tracepoint_works proc (in pending.exp) exits early as (according
> to the comment) the target was unable to install the fast tracepoint.
> 
> Could you confirm that the 'tstart' test immediately preceding this
> 'tstop' test passes successfully please, it's name will be:
> 
>   gdb.trace/pending.exp: ftrace works: start trace experiment
> 
> If this FAILs then I think, like the one above then it's OK that the
> corresponding tstop also FAILs.  If you are seeing the 'tstart' PASS
> then could you paste in the corresponding gdb.log snippet and I'll see
> if I can spot the problem.

The "start trace experiment" fails too, so it's probably the same thing
as the previous one.

Thanks for looking into it!

Simon