Re: Avoid "inline" after return type in function definitions

[Florian Weimer <fweimer at redhat dot com>]

* Joseph Myers:

> On Wed, 6 Feb 2019, Florian Weimer wrote:
>
>> Do you plan to submit patches to turn on individual -Werror= flags where
>> these are reasonable (such as -Werror=old-style-declaration)?
>
> I'm hoping it will be possible to use -Wextra with selected -Wno- options.  
> (Subject to care to avoid the use of -Wno- options for 
> nptl/tst-initializers1.c, which should get the full effect of -Wextra, 
> without -Wno-missing-field-initializers - whereas there are over a 
> thousand -Wmissing-field-initializers warnings building glibc, so I expect 
> -Wno-missing-field-initializers would be needed at least initially if 
> enabling -Wextra.)
>
> The warnings I'm using for the present fixes are from a single -Wextra 
> build for x86_64, of glibc but not the testsuite; actually adding new 
> warning options will need build-many-glibcs.py testing to find and fix 
> warnings more thoroughly.

My concern is that the dependency on -Wextra makes building glibc with
stock configure flags even more difficult than it is today.  -Wextra
seems to be some sort of dumping ground for things that aren't ready yet
for -Wall for various reasons, and most of the GCC developers I talk to
regularly strongly recommend not use -Wall -Werror (like we do).  -Wall
-Wextra -Werror would be worse.

Thanks,
Florian