Re: Inline definition problem in current

On Tue, Mar 24, 2009, Doug Barton wrote:
> David Schultz wrote:
> > On Mon, Mar 23, 2009, Gustau Perez wrote:
> >>  a few time ago I switched to current, right now I've it updated to 
> >> yesterday.  While compiling some ports (in fact, building x11/gnome2) I 
> >> found that some of them (written in C) are using  some inline functions 
> >> (I guess it is because the compiler will replace the call to the 
> >> function with the function itself). The problem is that gcc fails with 
> >> the following message :
> >>        
> >>        error: nested function 'XXX' declared but never defined
> >>
> >>   checking the code, the function is declared and then implemented in a 
> >> header file which is included in the offending .c file. The function is 
> >> declared as 'inline'. The only solution I found is to change the 
> >> definition to static.
> >>
> >>   Checking pontyhat shows me that many ports are failing because of 
> >> this problem. What I can understand is why is this happening, because 
> >> the same ports compiles fine in STABLE and the compilers's version in 
> >> base seems to be the same (gcc (GCC) 4.2.1 20070719  [FreeBSD], the same 
> >> in current) 
> > 
> > Which other ports were broken for this reason?
> 
> I am trying to compile gimp on -current right now and x11/babl and
> graphics/gegl both have this problem. Take a look at
> http://pointyhat.freebsd.org/errorlogs/i386-8-failure.html for more
> examples (click on the link on the right under Package to see the
> logs). There are currently over 600 broken ports in -current, all the
> ones I clicked on in a completely bogus sample had this same problem.

My bug; I missed an important line when merging from gcc trunk 122565.
Instead of reporting:

    error: nested function 'foo' declared but never defined

gcc should have been reporting:

    warning: inline function 'foo' declared but never defined

I'll check in a fix as soon as I run a buildworld.