Re: Re: [austin] Re: Defect in XSH Interaction of File Descriptors and Standard I/O Streams

[Joerg Schilling <yyyyyyyyy@xxxxxxxxxxxx>]

>From: David Korn <yyy@xxxxxxxxxxxxxxxx>

>> flushing an *input* stream is idiotic nonsense. To flush means to
>> pass buffered data forward to an operating system, it does not
>> mean discarding arbitrary, unknown amounts of input. The semantics
>> of flushing and discarding are miles apart.

>The purpose of this requirement was not to cause data to be
>sent back to the operating system, but to cause the seek pointer
>to be positioned to the end of the data that had been read.

>For example, if the implementation of the head utility uses
>buffered I/O, the file pointer would be positioned at end
>end of the data that it read so that a script such as
>       (head 5>/dev/null;cat) < file
>would successfully behead the first 5 lines of file.

>There are standard I/O implementations that do this and ksh
>calls fflush() on the input stream before running a command
>to get it seek pointer positioned correctly.

.... and if you have a stream open for reading and writing, you need
to call fflush() before switching the I/O direction.



Jörg

 EMail:yyyyy@xxxxxxxxxxxxxxxxxxxxxxxxxxx (home) Jörg Schilling D-13353 Berlin
       yy@xxxxxxxxxxxxxxx               (uni)  If you don't have iso-8859-1
       yyyyyyyyy@xxxxxxxxxxxx           (work) chars I am J"org Schilling
 URL:  http://www.fokus.gmd.de/usr/schilling   ftp://ftp.fokus.gmd.de/pub/unix