Re: fseeko bug

[Larry Jones]

Eric Blake writes:
> 
> According to Larry Jones on 12/14/2007 12:25 PM:
> > Eric Blake writes:
> >> maybe it's at least worth adding a compile-time assertion that 
> >> sizeof(off_t)
> >> ==sizeof(long) when !HAVE_FSEEKO.
> > 
> > Since I actually have a platform where that's not the case, I'd prefer a
> > run-time test that only objects when there's an actual problem:
> 
> Which platform?

BSD/OS 4.2

> And is there a ChangeLog entry for your patch?

2007-12-14  Larry Jones  <address@hidden>

        * lib/fseeko.c (rpl_fseeko): Don't refuse to compile when no native
        fseeko and off_t bigger than long but fail if the offset is too big.

> Does the platform lseek actually respect off_t?

Yes, it does.

> I still find it strange that there would not be an fseeko when the
> platform is capable of larger-than-long file offsets.

As I recall, POSIX had already defined lseek using off_t prior to the
Large File Summit that defined ftello and fseeko, so it was a simple
matter to define off_t as an extended type (quad_t), but there wasn't
any simple fix for the stdio library.

> But given your claim, it looks like you are correct that we
> are stuck with adding overflow checking in that case, and likewise for
> ftello.

Well, like I said before, BSD/OS, although very popular at one time, is
now quite thoroughly defunct, so I wouldn't go to too much trouble just
for it.  I don't mind not having working large file support, but I'd
like to be able to at least build code that uses gnulib fseeko and have
it work on regular files.

> But can we at least avoid compiler warnings for platforms where
> off_t is the same size as long, such that this condition can never be true?
> 
> > +  else if (offset < LONG_MIN || offset > LONG_MAX)

I don't know of any good way to do that since you can't use sizeof in
the preprocessor and there's no standard limit macro for off_t that I
know of.  Suggestions?

-Larry Jones

I take it there's no qualifying exam to be a Dad. -- Calvin