>> Well, the problem is how do you handle partial writes? Is there a
>> case in which an error occurs but the caller would want to know how much
>> data was successfully written? If not, you could just change the
>> interface to return a boolean 1 (all data successfully written) or 0
>> (error), though it would obviously be better to give such a routine a
>> different name and change the current code over manually. Otherwise, it
>> should probably act like a write() that doesn't return EINTR: i.e., if an
>> error occurs after writing some data out, it should return the number of
>> bytes successfully written, not -1.
>
>Personally I've found that knowing about a partial write is useless.
>If you only wrote part of it, there was probably an error that you
>want to report anyways. That's the whole point of a wrapper like
>pwrite which continues on meaningless errors (EINTR) but stops on
>larger errors.
That would be my own take as well (and is how I've written similar
routines in some of my personal projects), but as I'm sure you realize
after going through the transcode source, there's no telling what some
of those modules will do...
--Andrew Church
[EMAIL PROTECTED]
http://achurch.org/
