RFR 8027059: (sctp) fatal warnings overly restrictive with gcc 4.8.1

[Magnus Ihse Bursie]

On 2013-10-22 20:44, Chris Hegarty wrote:
> The SCTP native code complies with -Werror. gcc 4.8.1 reports unused 
> parameters as warnings, and this causes the build to fail. gcc 4.8.1 
> is the defaul on Ubuntu 13.10.

An alternative solution is to keep the warning but fix the code.

The traditional solution is to cast the unused variable/parameter to 
void, typically accompanied with a comment describing why we keep it 
even though it is unused, e.g. something like
(void) klass; // this is a formal JNI parameter which we don't use

However, it can very well be argued if this is indeed better than just 
disabling the warning. After all, it's a question of how to balance 
different kinds of code quality. If we have a lot of places were formal 
parameters like this, it might lead to worse code quality by littering 
it with void casts.

I just wanted to point out an alternative, I'm not protesting against 
your solution.

/Magnus

>
> These parameters are required as the functions are JNI functions. The 
> simplest solution is to ignore/suppress these unused param warnings.
>
> See discussion on:
> http://mail.openjdk.java.net/pipermail/build-dev/2013-July/009513.html
>
> diff --git a/makefiles/lib/NioLibraries.gmk 
> b/makefiles/lib/NioLibraries.gmk
> --- a/makefiles/lib/NioLibraries.gmk
> +++ b/makefiles/lib/NioLibraries.gmk
> @@ -150,7 +150,7 @@
>
>    ifneq ($(OPENJDK_TARGET_OS), macosx)
>
> -    SCTP_WERROR := -Werror
> +    SCTP_WERROR := -Werror -Wno-error=unused-parameter
>      ifeq ($(OPENJDK_TARGET_CPU_ARCH), ppc)
>        SCTP_WERROR :=
>      endif
>
> With this change it may be possible to remove the special casing of 
> ppc? I would need to check, but this could be handled separately.
>
> -Chris.