Re: improving manywarnings

[Simon Josefsson]

Eric Blake <address@hidden> writes:

> I'm considering using manywarnings in m4, and noticed that newer gcc
> supports even more warnings.  I noticed several warnings supported by gcc
> 4.3.4 --help=warnings that were not listed already; see the patch below.
> However, looking at it closer, it is obvious that some of the newly-added
> warnings added in this patch are C++ or ObjC-specific; maybe we should
> break this macro into the set of warnings useful for each language, rather
> than an overall list.  For that matter, some warnings (like -Wtraditional)
> don't make any sense in the context of gnulib, since we require C89 or
> better, so including them in gl_MANYWARN_ALL_GCC just means that package
> maintainers have to exclude it.

Yes, I've intentionally kept all non-C warnings out of manywarnings
because (in general, i.e. for some warnings) it may even lead to
problems when passed to gcc when compiling C.  So I don't think your
patch is likely to work well in normal situations.

I think it makes sense to keep separate lists, would you want to work on
it?

I'd prefer to keep a list covering all gcc warnings, so that I have to
explicitly anything from it.  I'd be fine with an approach that would
add -Wtraditional to another list, e.g. gl_MANYWARN_RECOMMENDED_DISABLED
and make it easy to disable all warnings from such a list.

/Simon