Re: [cp-patches] [Patch] JNI method checking during build time

[Michael Koch]

Am Montag, 15. November 2004 20:20 schrieb Tom Tromey:
> >>>>> "Michael" == Michael Koch <address@hidden> writes:
>
> Michael> I have written the attached patch to check at build time
> if Michael> all needed JNI methods are implement and no
> old/obsolete JNI Michael> method exists anymore.  This makes the
> build fail when the Michael> JNI methods get borked.
>
> I think this is a great idea.
>
> Michael> Ok to commit ?
>
> I think the script is great, but see below.
>
> Michael> +noinst_DATA = check_jni_methods
> Michael> +check_jni_methods:
>
> It is better to use all-local for this sort of thing.
> You might also want to consider how this interact with rebuilding
> the headers.  It probably makes more sense to put the checking rule
> in include/Makefile.am and have it depend on $(H_FILES).

Thanks for the hint about 'all-local'. I don't knew this target 
before. I will use it now.

I thought about where to put the check before I have done it. Your 
solution will only do the check when one of the headers change and 
not when the c files change. Its possible that developers add a 
method to a c file and forget to regenerate the header file. This 
script makes it fail when this happens. When it depends on $(H_FILES) 
in include/Makefile.am this will not happen.

What do you think ?


Michael
-- 
Homepage: http://www.worldforge.org/