[Cocci] "TODO: FunCall" in python match printing
Yoann Padioleau
padator at wanadoo.fr
Fri Mar 6 16:50:19 CET 2009
Julia Lawall <julia at diku.dk> writes:
>> > For some other files, the warning is different:
>> > (ONCE) ast_to_flow: filter a directive
>>
>> Ok, this one is mine ... it tells you that the engine
>> will not represent some cpp directive like #ifdef or #define instructions
>> in the internal structure of coccinelle. You should not be concerned
>> by that, it's mainly for the developer of coccinelle.
>>
>> (Julia, maybe we can do pr2_developer/pr2_user ? or maybe just put
>> those pr2 behind a flag).
>
> Still the user might be interested to know that some #defines are not
> being treated.
Yes, but I think those messages are mostly issued for some #ifdef lines
inside the body of functions.
> Although the message is not very helpful...
As there are lots of things we still don't handle, and as I think most
of them are not harmful regarding most SPs (those code are ifdef and
SmPL does not handle ifdef directives anyway, or GetRefLabel and again
SmPL does not handle that too), and as apparently each time the user
wonder what we mean, I think it's better for all those to not say
anything in "user" mode. They will review the code anyway. And
basically each message really say "hey, because of some limitations,
they are problably code that we don't process , so maybe we are not
complete". So we could just achieve the same effect by just saying
nothing. I think people know our tool is not perfect ...
I don't think Coverity says "hey, we processed only X% of your
code and there are lots of things we don't handle".
Or maybe we just don't say anything during the run, and just
at the end we could say:
"we were not always able to type all expressions,
probably because we didn't find all the necessary header files.
In most cases this is not a problem, but in doubt you can
check for the -all_includes, or -xxx, or -zzz spatch options".
But really I think we sould say nothing, and put instead those
comments in the manual.
>
> julia
More information about the Cocci
mailing list