_F64_ functies

_F64_ functies

Post by Stefan Smeet » Wed, 06 Nov 2002 21:51:39



Can someone tell me where I can find the functions in Tru64 Unix 4.0F:
_F64_lstat
_F64_stat
_F64_fstat

In which shared libraries are they included ?

Thanks,
Stefan

 
 
 

_F64_ functies

Post by Brian Hale » Thu, 07 Nov 2002 01:04:00



> Can someone tell me where I can find the functions in Tru64 Unix 4.0F:
> _F64_lstat
> _F64_stat
> _F64_fstat

Looks to me like you're trying to run a binary compiled on a V5.X system
on a V4.X system.  You can't do that.

-Brian

 
 
 

_F64_ functies

Post by Bob Harri » Thu, 07 Nov 2002 07:58:49




X Can someone tell me where I can find the functions in Tru64 Unix 4.0F:
X _F64_lstat
X _F64_stat
X _F64_fstat
X
X In which shared libraries are they included ?
X
X Thanks,
X Stefan

In v5.0 we provided larger fields in the stat (and statfs) structures.
This required new APIs _BUT_ we still had a requirement to still support
the OLD APIs so that existing applications could continue to run.  
Though a lot of complex #ifdef tricks in the stat.h (and mount.h) header
files, and lots of work in libc as well as the kernel, we were able to
provide the old and new stat (and statfs) APIs in such a way that users
would not need to know about it and allow existing v4.* and earlier
binaries to continue to operate, and for v5.0 to pass all the standards
tests (not an easy trick at all, let me tell you).

However, any program compiled on v5.0 or later is not going to run on a
v4.* system.  The only supported work around would be to compile the
code on a v4.* system and then use that on both the v4.* system and the
v5.* system (that would be supported).

                                        Bob Harris

 
 
 

_F64_ functies

Post by Mike Colema » Fri, 08 Nov 2002 13:50:42



> Though a lot of complex #ifdef tricks in the stat.h (and mount.h) header
> files, and lots of work in libc as well as the kernel, we were able to
> provide the old and new stat (and statfs) APIs in such a way that users
> would not need to know about it and allow existing v4.* and earlier
> binaries to continue to operate, and for v5.0 to pass all the standards
> tests (not an easy trick at all, let me tell you).

You're the one!  

This stuff broke gcc since it wasn't defining stat as a function, which is
required by those standards, AFAIK.  (I think this is now fixed (i.e., worked
around) in the latest version of gcc.)

And don't even get me started on gdb...

Mike

 
 
 

_F64_ functies

Post by Bob Harri » Sat, 09 Nov 2002 03:09:46





X > Though a lot of complex #ifdef tricks in the stat.h (and mount.h) header
X > files, and lots of work in libc as well as the kernel, we were able to
X > provide the old and new stat (and statfs) APIs in such a way that users
X > would not need to know about it and allow existing v4.* and earlier
X > binaries to continue to operate, and for v5.0 to pass all the standards
X > tests (not an easy trick at all, let me tell you).
X
X You're the one!  

I guess so.

X This stuff broke gcc since it wasn't defining stat as a function, which is
X required by those standards, AFAIK.  (I think this is now fixed (i.e., worked
X around) in the latest version of gcc.)
X
X And don't even get me started on gdb...
X
X Mike

The most recent stat.h does not use #define for non-DECC compilers, but
instead uses jacket routines.  This change went in about a year ago.  As
far as I know this was the last reported problem against stat.  If there
are still problems, they should be escalated via an IPMT with the
Customer Support Center.

                                        Bob Harris