svn commit: r187805 - head/lib/libc/stdio

[Giorgos Keramidas]

On Wed, 28 Jan 2009 09:07:51 -0500, John Baldwin <jhb at freebsd.org> wrote:
> On Wednesday 28 January 2009 8:55:37 am Tom Rhodes wrote:
>> On Wed, 28 Jan 2009 15:09:56 +0200 Giorgos Keramidas <keramida at ceid.upatras.gr> wrote:
>> > On Wed, 28 Jan 2009 01:11:21 +0000 (UTC), Tom Rhodes <trhodes at FreeBSD.org> wrote:
>> > > Author: trhodes
>> > > Date: Wed Jan 28 01:11:20 2009
>> > > New Revision: 187805
>> > > URL: http://svn.freebsd.org/changeset/base/187805
>> > >
>> > > Log:
>> > >   Remove comment about clearerr() being the only method of clearing
>> > >   the EOF indicator, fseek() may also be used for this.
>> > >
>> > >   Bump document date.
>> >
>> > I don't like this, sorry...  Having a pointer to clearerr() is nice.
>> > Removing it *deletes* useful information, but we should add _more_ of
>> > it.
>> >
>> > How about this instead?
>> >
>> >     The end-of-file indicator may be cleared by explicitly calling
>> >     .Fn clearerr ,
>> >     or as a side-effect of other operations, i.e.\&
>> >     .Fn fseek .
>>
>> I think 'side-effect" is wrong here - it may not be a "side
>> effect" at all, but, on purpose.  :)
>
> If one solely wants to clear the indicator then clearerr() is probably
> what you should do.  Using fseek() only to clear the indicator would
> be bad form.  One should be using fseek() because they need to seek to
> a different location in the stream, not to clear the error.  Thus, I
> agree with Giorgos' wording.

Precisely.  We are not suggesting that users SHOULD use side-effects,
just noting one example.  The _intent_ of a function to clear EOF is
more important than the fact that it happens 'in addition to' other
things as opposed to 'because we asked for it'.