On 9/25/10 5:07 PM, Mats Kindahl wrote:
> On 09/25/2010 03:31 AM, Davi Arnaut wrote:
>> On 9/24/10 12:22 PM, Mats Kindahl wrote:
>>> In this particular case, we would probably be better off if we could
>>> turn on -std=c89, which disallows C++ comments in C code.
>>>
>>> The downside is that it would of course mean that we have to start
>>> writing standards-compliant C code (oh horrors!). :)
>>
>> I wish that overtime we could focus more on styles and actions, such
>> as this, that have more practical consequences. Over time, discussions
>> about the cosmetic aspects of coding style tend to get very silly.
>
> Ah, yes... :)
>
>> How about a proposal that new C and C++ code must be
>> standards-compliant? That it must compile warning free? etc.
>
> Oh, I would love that, but I wonder how that can be handled practically.
> New code in separate files? Rewriting the files one by one?
Fixing issues one by one.
> It would be great if -ansi or -Werror could be enabled on a per-file basis.
We already do -Werror for all files. See the maintainer mode option:
C_WARNINGS "-Wall -Wextra -Wunused -Wwrite-strings -Werror"
CXX_WARNINGS "${MY_MAINTAINER_C_WARNINGS} -Wno-unused-parameter"
but this side steps the strict aliasing warnings as we have too many
violations of it. It will take some time to fix them all.
I'm planning to submit a style request that we enable the maintainer
mode by default for debug builds.
>> Do you have an idea of how much work is needed to make MySQL compile
>> with -std=c89?
>
> Sadly, yes. I did make an attempt to rewrite some aspects I disliked
> (the cast of string literals to "char*", it didn't look like much work),
> but it turned out to ripple through the code. If we could handle it on a
> one-file-at-a-time basis, I think it could be done.
Actually, I just tested HAVE_CMAKE=no
BUILD/compile-pentium-valgrind-max-no-ndb --warning-mode=pedantic in 5.5
and it built fine after fixing one case of C++ style comment in C file.
It builds C files with -ansi -pedantic and C++ ones also with -std=c++98.
But there were quite a few warnings. I guess the challenge is to get it
working without warnings (-Werror). For this we would need proper
configure time checks for both cmake (like, mapping inline to
__inline__, likewise for asm) and working around the silly stuff such as
%p requiring the argument to be of type void pointer (non-standard, but..).
Will take a more detailed look at the warnings later. Anyway, quite nice
that it actually compiles successfully.
Regards,
Davi
