On 21/4/2022 5:45 pm, Karel Gardas wrote:
> On 4/21/22 02:12, Chris Johns wrote:
>> On 21/4/2022 4:38 am, Karel Gardas wrote:
>>>
>>> Chris, Frank,
>>>
>>> I don't feel particularly comfortable with this patch. It works here on
>>> Ubuntu
>>> 20.04 with python 3.3 and Ubuntu 22.04 with Python 3.10, but I have to admit
>>> semantics of this cfg language is beyond my liking and a lot is in complete
>>> darkness here. E.g. how &&, < operator are working here. Delete both int
>>> conversions and it'll not work at all. Keep one and it'll work well.
>>> The problem is probably lack of time to go thorough all the docs on
>>> docs.rtems.org to find relevant bits while working on the patch.
>>>
>>> Anyway, if you have time, please review and test on your platforms.
>>
>> I think you are right to be concerned. I cannot find in the parser any
>> support
>> for `int()` which makes me wonder if `int(3)` is being compared but I am not
>> sure.
>
> Hmm. And I've thought cfg format is just DSL interpreted in python hence
> translated internally to python hence my int() usage.
>
>> Leave this with me and I will take a closer look.
>
> Absolutely!
>
>> I think we will need to add a
>> numerical compare. Hmmm.
>
> Another solution may be to enhance version string to print every number with
> two
> digits and with leading zero if needed and concatenate string together. This
> may
> probably work too even with lexi comparison.
> E.g.
>
> 2.7.1 -> "020701"
> 2.7.17 -> "020717"
> 3.9.2 -> "030902"
> 3.10.3 -> "031003"
>
> Hard to judge which approach is cleanest one.
This breaks when we exceed 99 etc. Maybe that will happen, maybe it will not.
I wonder if hacking the parser to handle this provides a more long term solution
beyond this use case.
The problem becomes adding support to the language. I wonder if checking if both
the left and right hand sides in a compare are numeric, ie 3 or 3.0, and if they
are make the comparison numeric? To force a lexicographical comparison simply
have `%if x%{yyy} < 3.8`?
Chris
_______________________________________________
devel mailing list
devel@rtems.org
http://lists.rtems.org/mailman/listinfo/devel
