Re: [PATCH v6 5/8] GDB/Python: Use None for `var_zuinteger_unlimited' value set to `unlimited'

[Luis Machado <luis.machado@arm.com>]

On 10/31/22 13:08, Maciej W. Rozycki wrote:
> On Mon, 31 Oct 2022, Luis Machado wrote:
> 
>> A bisect confirmed commit e7e1f2034567207e5e01cb75ea2ffd568a64e84d changed the
>> gdb.python/py-parameter.exp behavior from PASS to FAIL.
>> gdb.python/py-format.exp was failing before.
>>
>> I've attached the relevant gdb.log and gdb.sum.
>>
>> I don't fully understand the issue here, so I can't be sure this is a real
>> regression or something that was broken before but this patch just happened to
>> uncover it.
> 
>   Thanks.  There is this error:
> 
> Traceback (most recent call last):
>    File "<string>", line 5, in <module>
>    File "<string>", line 4, in __init__
> RuntimeError: Range exceeded.
> Error while executing Python code.
> (gdb) PASS: gdb.python/py-parameter.exp: test_integer_parameter: kind=PARAM_UINTEGER: create parameter: input 7: end

Yeah, I found this to be a bit odd.

> 
> in creating the parameter earlier on and consequently all the subsequent
> tests on the parameter fail.  I think the case above shouldn't score as a
> PASS either; pattern matching must be too liberal here.
> 
>   The situation here is previously only parameters of the PARAM_ZUINTEGER
> and PARAM_ZUINTEGER_UNLIMITED types were tested.  Now we also test ones of
> the PARAM_UINTEGER, PARAM_INTEGER, and PARAM_ZINTEGER types, so it is a
> new test that is failing.  Parameter of all the types are initialised to
> 0, which is permitted either as itself or as the "unlimited" value.  For
> the failing PARAM_UINTEGER case it means "unlimited".  So wonder where
> this range error is coming from really, hmm...

Thanks for the details.

> 
>   You don't happen to have any local patches applied, do you?  Also what
> version of Python?

No. This is a top-of-tree build. This happens for both Ubuntu 20.04 (Python 3.8.10) and Ubuntu 22.04 (Python 3.10.4).

I have to say I noticed, in the past, some variability in the results of these python tests due to version. But in this
case it seems to be a reliable outcome.

I can try to pinpoint where that range exceeded message is coming from. It might give us answers.

> 
>    Maciej