At 09:39 AM 4/30/2001 -0400, Henderson, Jordan wrote:
> > From: Dan Sugalski [mailto:[EMAIL PROTECTED]]
> > At 09:24 AM 4/30/2001 -0400, [EMAIL PROTECTED] wrote:
> > >"Craig A. Berry" <[EMAIL PROTECTED]> writes:
> > > > I think this is a reproducer for the problem that occasionally
> > > > inserts spurious carriage returns into what should be one line of
> > > > output in the test suite. It appears to be a function of
> > pipe buffer
> > > > size. If the mailbox buffer for a pipe fills up, reading EOF from
> > > > the pipe causes a carriage return to get inserted in the stream.
> > >
> > >Well, much of this is from the interaction between the CRTL and
> > >mailboxes. There's a limit to what we can do to wedge our fixes in
> > >between those two. There's a definite limit to what I'm willing to do
> > >to rewrite the CRTL i/o library.
> >
> > Would it be better, then, to drop to $QIO calls for this
> > then? We could
> > wedge that in reasonably easily, as we do have a goodly
> > amount of control
> > at this level.
>
>Isn't dropping to $QIO calls to do these functions essentially the same as
>rewriting the CRTL i/o library?
To some extent, yes, but only in a very limited set of circumstances. I'm
not considering rewriting the CRTL stdio library (well, OK, I am, but
that's a separate issue), just skipping it for communication with our child
process.
>I'm constantly frustrated by the odd assumptions that I can only test by
>experimentation made in the Compaq CRTL libraries. I really think a lot of
>confusion could be cleared up if the Compaq CRTL libraries were available as
>Open Source.
No, that'd make it worse. Right now any undocumented behaviour's undefined,
and subject to change without notice. (Which it does) If you had the source
available, it'd be easier to see what the CRTL does (likely accidentally or
wrong) in the undocumented corner cases. Experimentation sucks too, but at
least when they break things that way they can say "we didn't document it"
>But, whenever I suggest that I get nothing but derisive comments, complaints
>about the chaos that would result and other speculations.
>
>The primary goal of having a CRTL in source form that I can study and
>experiment with is not to proliferate versions, but to understand and work
>WITH what is provided. Also, sensible modifications could be submitted
>back to Compaq for potential inclusion in future releases.
I have the source, and I can honestly say I don't think it'd do all that
much good. That's just me, though. (I admit I've ben tempted to take a
whack at it myself, to get fdopen working on sockets if nothing else...)
Dan
--------------------------------------"it's like this"-------------------
Dan Sugalski even samurai
[EMAIL PROTECTED] have teddy bears and even
teddy bears get drunk
