Re: PCI: Remove "extern" from function declarations

Re: PCI: Remove "extern" from function declarations

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 1099 bytes --]

Hi Bjorn,

I just noticed this commit (f39d5b72913e "PCI: Remove "extern" from
function declarations") in the pci next branch.  Just wondering why you
would do such a thing.  It is a lot of churn in quite a few header files
and (in my opinion) goes the wrong way anyway.

It may not actually make any difference to the compiler for functions, but
for variables, it does.   A variable declared in a header file without
"extern" will effectively define it in every compilation that includes
the header file, one with "extern" will only produce references.

So, since, the global variables really should have the "extern", the
functions are now inconsistent with that.

It also means that when someone copies a function declaration to make a
"static inline" stub, the "extern" is a big hint for them to remember to
make the stub "static" - I have seen several occasions when this was not
done and that causes compilation failures (but usually only in the
configuration that the author did not bother to test).
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au

[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]

Re: PCI: Remove "extern" from function declarations

[Bjorn Helgaas]

On Wed, Apr 17, 2013 at 5:24 PM, Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> Hi Bjorn,
>
> I just noticed this commit (f39d5b72913e "PCI: Remove "extern" from
> function declarations") in the pci next branch.  Just wondering why you
> would do such a thing.  It is a lot of churn in quite a few header files
> and (in my opinion) goes the wrong way anyway.
>
> It may not actually make any difference to the compiler for functions, but
> for variables, it does.   A variable declared in a header file without
> "extern" will effectively define it in every compilation that includes
> the header file, one with "extern" will only produce references.
>
> So, since, the global variables really should have the "extern", the
> functions are now inconsistent with that.
>
> It also means that when someone copies a function declaration to make a
> "static inline" stub, the "extern" is a big hint for them to remember to
> make the stub "static" - I have seen several occasions when this was not
> done and that causes compilation failures (but usually only in the
> configuration that the author did not bother to test).

Good points.  The fact that some function declarations use "extern"
and some don't is just an annoying inconsistency.  I prefer fewer
words to read, so I removed them, but that's just a personal
preference, and I'd be OK it were always present, too.  I suppose if
anybody but me actually cared very much, we'd have checkpatch complain
about one way or the other, but I don't think it does.

Bjorn