Re: pkg/45248 (net/samba35 build fails on i386-5.99.55 (201108122100Z))

[David Holland <dholland-pbugs%netbsd.org@localhost>]

The following reply was made to PR pkg/45248; it has been noted by GNATS.

From: David Holland <dholland-pbugs%netbsd.org@localhost>
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: pkg/45248 (net/samba35 build fails on i386-5.99.55
 (201108122100Z))
Date: Sun, 14 Aug 2011 04:17:38 +0000

 On Sun, Aug 14, 2011 at 01:35:02AM +0000, Emmanuel Dreyfus wrote:
  >  I see: I guarded the new functions by #ifdef to make sure they would not
  >  be visible to programs unaware it was not finished, but I did not though
  >  about GNU configure's way of discovering things that are not defined in
  >  the headers.
 
 Right... you've been autoconfiscated.
 
  >  There are multiple fixes possible:
  >  
  >  1) GNU libc defines macros to spot unimplemented functions, and GNU
  >  configure uses that feature to spare them. We can quickly workaround
  >  Samba build failure by adding in pkgsrc/net/samba35/Makefile
  >  CFLAGS+=        -D__stub_utimensat
  >  This is should work arround the build problem immediatly. Can you let me
  >  know if it does?
 
 A cleaner way is to set ac_cv_func_utimensat=no in CONFIGURE_ENV.
 
 The question I guess is what version number to attach it to...
 
  >  2) #define __stub_utimensat in <limits.h> to obtain the same effect
  >  system-wide.
 
 That's awfully gross... but it does I guess deal with this problem for
 other packages as well.
 
  >  3) Samba does not use the file descriptor argument of utimensat(),
  >  therefore we can quickly implement enough of it for Samba to use it in
  >  -current, it does not depend on namei upgrade. This is just a utime able
  >  to return struct timespec.
 
 Maybe this is the best route though.
 
 (Also, the namei changes are no longer a serious problem...)
 
 -- 
 David A. Holland
 dholland%netbsd.org@localhost