Re: [committed][gdb/testsuite] Make gdb.mi/mi-sym-info.exp more robust against timeouts

[Luis Machado via Gdb-patches <gdb-patches@sourceware.org>]

On 4/30/21 10:28 AM, Simon Marchi wrote:
> On 2021-04-30 9:20 a.m., Luis Machado wrote:> On 4/30/21 10:06 AM, Simon Marchi wrote:
>>> On 2021-04-30 8:55 a.m., Luis Machado via Gdb-patches wrote:
>>>> Wouldn't it be more reliable to do line-by-line matching instead of sprinkling timeouts across the test? Or so I've been told.
>>>>
>>>> This doesn't seem to be an instance where GDB is taking too long, but rather that the MI output is very verbose, right?
>>>
>>>   From what I can see, it's a case of GDB thinking for a long time
>>> (gathering symbols) before outputting anything.  You can see it
>>> by having a
>>>
>>>     $ tail -F testsuite/gdb.log
>>>
>>> on the side, while you run the test.
>>>
>>> Another one that I found times out in a similar way on my machine is
>>> gdb.base/info-os.exp.
>>>
>>> Simon
>>>
>>
>> Fair enough. But I also see the output being 113K chars in one case, no line breaks. I think these instances of very long lines are something to be dealt with.
> 
> Yeah, good point.  If GDB takes 6 seconds to think and Dejagnu takes 6
> seconds to read the long input, the timeout is a result of the
> combination of both, so it wouldn't hurt to parse item by item.  It's a
> bit more work to parse one item at the time, but in the end I think it
> also helps when you need to debug the test, as you can print a little
> something after parsing each item and know where things went wrong.

What I'm unsure about is if the MI protocol cares about line breaks or 
not. If not, there is no reason we should be putting out such long line, 
as it is hard on the test infrastructure.

> 
> Simon
>