[OpenAFS-devel] rxk5 branch is ready; please test

[Douglas E. Engert]

Russ Allbery wrote:
> Jeffrey Hutzelman <jhutz@cmu.edu> writes:
>> "Douglas E. Engert" <deengert@anl.gov> wrote:
>>
>>> The correct way for AFS is it test for HAVE_FUNCTION_MACRO which
>>> configure will set if the compiler can handle __FUNCTION__. Then again
>>> the new rxk5/servconn.c is the only source routines in AFS that uses
>>> __FUNCTION__ as far as I can tell.
>> I think we're talking about two different issues.
>>
>> Yes, obviously we shouldn't use __FUNCTION__ if it isn't available.
>> Unfortunately, I'm pretty sure it's _not_ a preprocessor macro, which
>> means you can't test for it with #ifdef -- someone will have to write a
>> real test to see if it's there.  Or we could just avoid using it.

There is a real test, src/cf/function-macro.m4 and AFS configure
will call it and set HAVE_FUNCTION_MACRO. So far AFS has avoided using
__FUNCTION__, accept in this one case in new test code where it is not really needed.


> 
> __FUNCTION__ is a gccism.
> 
> /* __func__ is C99, but not provided by all implementations. */
> #if __STDC_VERSION__ < 199901L
> # if __GNUC__ >= 2
> #  define __func__ __FUNCTION__
> # else
> #  define __func__ "<unknown>"
> # endif
> #endif
> 
> works for pam-krb5 and pam-afs-session.
> 
>> But my point wasn't about "how do we deal with __FUNCTION__ not
>> existing", or even "what should we use here".  It was lower-level than
>> that -- if what you want to do is concatenate __FILE__ and a literal
>> string, the way to do so is with juxtaposition.
> 
> This may not work in all cases since, for whatever reason, some compilers
> implement __func__ or __FUNCTION__ as a weird static string instead of a
> literal and concatenation doesn't work.  Annoying.

Yes, its annoying. But the Solaris compiler is used to compile OpenAFS.
> 

-- 

  Douglas E. Engert  <DEEngert@anl.gov>
  Argonne National Laboratory
  9700 South Cass Avenue
  Argonne, Illinois  60439
  (630) 252-5444