Re: bleadperl: PerlIOStdio_invalidate_fileno breaks fclose semantics

Michael Schroeder <Michael.Schroeder@informatik.uni-erlangen.de> writes:
>Hi porters,
>
>the fix of the socket memory hole doesn't take into account that
>fclose() flushes the stdio stream. This flushing doesn't work anymore
>if fileno gets invalidated. As a result most web services will cease
>to work.

Hmm, there used to be code to fflush() in there - will take another 
look.

>
>Test program:
>
>open(T, ">/tmp/test") || die;
>open(D, ">&=".fileno(T)) || die;
>print D 'ok';
>close D || die;
>close T || die;
>open(T, "</tmp/test") || die;
>my $r = <T>;
>close T;
>$r ||= 'fail';
>print "$r\n";
>
>Cheers,
>  Michael.
-- 
Nick Ing-Simmons
http://www.ni-s.u-net.com/

• bleadperl: PerlIOStdio_invalidate_fileno breaks fclose semantics by Michael Schroeder
• Re: bleadperl: PerlIOStdio_invalidate_fileno breaks fclose semantics by Nick Ing-Simmons
• Re: bleadperl: PerlIOStdio_invalidate_fileno breaks fclose semantics by Abhijit Menon-Sen
• Re: bleadperl: PerlIOStdio_invalidate_fileno breaks fclose semantics by Nick Ing-Simmons