On Tuesday 14 November 2006 19:00, Darryl Gregorash wrote:
> On 2006-11-14 17:20, Theo v. Werkhoven wrote:
> > Tue, 14 Nov 2006, by [EMAIL PROTECTED]:
> >
> >
> > <snip>
> >
> >> The end result must be a file, and I do not know if a named pipe
> >> qualifies as a file in mozilla's language. If it does, then I can
> >> just insert "sig" as the signature file, but I am not prepared to
> >> go through the exercise. I will leave that to Hylton, if he is
> >> interested.
> >
> > So will I.
>
> Well, I tried it anyway.. trying to read a named pipe in Seamonkey
> just hangs.

Pipe semantics are perhaps a bit odd. Readers block if there's at least 
one writer present (i.e., a process that has the pipe open for 
writing). If there are no writers, readers get and EOF (reads 
immediately return a count of 0 bytes read). If a writer writes when 
there are no readers present, the write is discarded and the writing 
process is sent a SIGPIPE. Writers will block once they reach the 
high-water mark for pipes, which is something in the 8-16 K mark (I'm 
guessing--it may be higher these days). If there are multiple writers, 
there are no guarantees about relative ordering of concurrent writes. 
Likewise for multiple readers, but no byte written will be read by more 
than one reader.

If you want to use pipes, you need to keep these things in mind. Named 
pipes are no different than conventional ones, except that the reader 
and writers need have no common process parentage from which to have 
inherited the open pipe descriptors.


Randall Schulz
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to