[PATCH 2/2] libio: Perform write-only fflush as part of freopen (bug 30731)

Florian Weimer fweimer@redhat.com
Tue Aug 8 13:59:37 GMT 2023


* Andreas Schwab:

> On Aug 08 2023, Florian Weimer via Libc-alpha wrote:
>
>> For standard streams that have a file descriptor, the read flush
>> may change the file offset of the underlying file descriptor.  This
>> is undeseriable because it makes using freopen after fork unsafe
>> even for read-only files.
>
> The POSIX standard says that fflush should be called unconditionally,
> which means that it expects all side effects to happen.

Hmm.  I think we might be in undefined territory here.  But taking a
second look at it, the current (unpatched) behavior is probably more
generally useful.

I would like to apply the first patch as a cleanup, though.

Thanks,
Florian



More information about the Libc-alpha mailing list