Hi,

Neil Jerram <neil@ossau.uklinux.net> writes:

> ludo@gnu.org (Ludovic Courtès) writes:
>
>> The attached patch adds a new `scm_t_off' type, whose definition does
>> not depend on the application's `_FILE_OFFSET_BITS' value.  Can you
>> confirm that it allows you to build Guile with 32-bit offsets and Snd
>> with 64-bit offsets (or vice versa)?
>>
>> Neil: Does this sound like the right approach?
>
> Yes, this looks great.  (In case it makes a difference, I've reviewed
> the master commit, not the patch that you put in email.)

Perfect, thank you!

> Also I think it means that we can delete some code that was needed to
> handle the possibility that off_t might be 32-bit on a platform that
> also supports large files.
>
> - In scm_seek (), the "if (SCM_OPFPORTP (fd_port))" block is now not
>   needed, because the following more general "if (SCM_OPPORTP
>   (fd_port))" case will handle 64-bit correctly.
>
> - Therefore scm_i_fport_seek () can be removed.
>
> - The "#if GUILE_USE_64_CALLS && HAVE_STAT64 && SIZEOF_OFF_T !=
>   SIZEOF_OFF64_T" implementation of fport_seek () can be removed,
>   because fport_seek () is now always identical to
>   fport_seek_or_seek64 ().
>
> - Plus the same things again but for "truncate" instead of "seek".

Also:

 - `scm_t_off' and `off_t_or_off64_t' are now identical, so the latter
   could be removed.

> Would you agree?

Yes.  What do you think of the attached patch?

On my GNU/Linux machine where 64-bit offsets are used, I see this:

--8<---------------cut here---------------start------------->8---
scheme@(guile-user)> (open-input-file "/dev/zero")
$1 = #<input: /dev/zero 70>
scheme@(guile-user)> (seek $1 (expt 2 33) SEEK_SET)
$2 = 0
--8<---------------cut here---------------end--------------->8---

I can't think of a better test case, though.

Thanks,
Ludo'.