Re: Bit field issues in Solaris 2.8

From: Tom Hughes (thh_at_cyberscience.com)
Date: Fri Aug 08 2003 - 14:28:12 BST

  • Next message: Matt Seitz: "Re: ups Next and Step acting like Cont"
    In message <1f299d1e4c.work_at_loxley.compton.nu>
            Tom Hughes <thh_at_cyberscience.com> wrote:
    
    > In message <20030808003726.GB50957_at_merton.aus.deuba.com>
    >           Callum Gibson <callum.gibson_at_db.com> wrote:
    > 
    > > On FreeBSD display/edit works fine. On Solaris, it displays zeroes and
    > > just gobbles up all values when edited. However, the edited values are
    > > stored because the printf shows correctly what was entered. This looks like
    > > a problem that was introduced in 3.38b1 (our previous Solaris version was
    > > 3.37) on Solaris (2.6 - the original poster was on 2.8). Perhaps an endian
    > > problem?
    > 
    > I suspect so, as I already told Steve. He says it is broken in 3.37 as
    > well though, and my tests on Solaris tend to agree. I will try and look
    > at this soon.
    
    It was an endianess problem, and 3.37 does work as far as I can, at
    least for bitfields based on 32 bit types. I don't believe it worked
    for bitfields based on 16 bit types however, which is what my original
    fix was all about (that and 64 bit bitfield support).
    
    Unfortunately my previous was wrong on big endian platforms. A fix has
    just been committed to CVS and a patch against CVS is attached.
    
    Tom
    
    -- 
    Tom Hughes (thh_at_cyberscience.com)
    Software Engineer, Cyberscience Corporation
    http://www.cyberscience.com/
    
    
    
    



    This archive was generated by hypermail 2.1.4 : Tue Aug 12 2003 - 15:54:30 BST