This is the mail archive of the
fortran@gcc.gnu.org
mailing list for the GNU Fortran project.
Re: [RFC] fseek, an alternative implementation (PR22359)
Tobias Schlüter wrote:
Brooks Moses wrote:
I think we're probably in full agreement that we should do the
simplest possible solution here that isn't fundamentally broken, with
the intent that we can close the PR about FSEEK not being implemented,
and hopefully never have to worry about this again. :)
My personal feeling -- and this is where I think we disagree -- is
that the absence of some way to indicate an error means that the
function is a little bit broken (which bothers me, albeit more on an
aesthetic than a practical level), and it's a pretty trivial amount of
additional complexity to add it. Daniel seems to be willing to
include it, so thus I'd say we should let him if he wants to.
On the other hand, if Daniel were to offer an implementation which
doesn't have STATUS implemented, I wouldn't object to that either.
Mainly I'm just saying that I disagree with saying that he should
remove it.
It's definitely not worth spending more time on this :) I take it that
you believe that adding a g77-compatible version will not be made harder
by this patch, and therefore retract my objection. I looked at the
patch, and except for the choice of the right integer kinds, as
mentioned in the other replies, it looks sound to me.
Just after I wrote this, I found an ambiguity Daniel's patch would
cause: imagine someone assigns a label to an integer variable, and
passes this as last argumetn to FSEEK. Sigh, looks like this issue
doesn't want easy solutions. Do as you want Daniel, I don't think this
an important objection and anybody who does this deserves to be bitten.
- Tobi
Cheers,
- Tobi