Re: qmail-1.03: any volunteers?!

["Robert Collins" <robert dot collins at itdomain dot com dot au>]

----- Original Message -----
From: "friedman_hill ernest j" <ejfried@california.sandia.gov>
To: <cygwin@cygwin.com>
Sent: Saturday, September 29, 2001 8:16 AM
Subject: Re: qmail-1.03: any volunteers?!


>
> Here's the begining of a simple mkfifo implementation for Cygwin.  I

Maybe I'm dense... but where is it? I can't see a patch or a url
reference in your email. I've also written a fifo implemention, feature
complete except for a distressing tendency to die on fork, which I
haven't looked at since february this year (and since then I've learnt a
_lot_ more about fork()'s guts. Therefore I think I'm in a position to
perform peer review for your "simple" implementation (Note: I have no
idea what you are implying by referring to it as simple). I assume you
mean somewhat featureless, as opposed to "very small and clear code".
Please correct my assumption!).

> did this about a year ago. I went so far as to get the paperwork
> approved to let me sign this over to Cygnus. I tried to get some help
> from this list on how to integrate it with Cygwin, but was met with

I don't recall this. So I did a search under "ejfried" of the cygwin and
cygwin-developers list archives...
I found http://sources.redhat.com/ml/cygwin/2001-02/msg00931.html which
was a reference to ORBS trouble, and Chris not wanting to discuss a fifo
implementation in private - which IMO is absolutely good common sense.
Mailing lists for developers are essential to keep a project syncronised
and working well, without too much duplicate effort being spent. I
certainly didn't see any

> sufficiently rude and unhelpful responses

to your emails - even though you apparently wrote your responses
off-list! (see for example a reply to an email from you, but your email
is no on the list archive...
http://sources.redhat.com/ml/cygwin/2000-09/msg00447.html).

> that I abandoned the project
> (I suspect this happens to quite a few folks who are interested in
> contributing, but don't have the free time to devote to untangling the
> Cygwin "kernel" source themselves. But I digress...)

Possibly. There are dedicated resources for developers - the developer
list, the cygwin internal documentation. See the recent thread on magic
dir listing for example. It certainly is a shame you ended up abandoning
it - I could have saved myself a bunch of time.

> You can't use Windows named pipes to implement UNIX fifos -- the
> semantics are just too different.

Aren't they just :].

> Anyway, anyone who would like to try their hand at implementing fifos
> may feel free to use this as a starting point.

I would be interested in comparing it with mine. Could you please post
the URL?

I've attached a (somewhat rusty) patch kit for my implementation. It's
currently broken - I started a little, errrm, virtual function bomb
approach when I didn't really grok what happens with HANDLEs and fork().
(As I mentioned, no time yet to revisit). Theres also a bunch of crud in
there from other partial projects. This working dir jsut recently got
put aside, and I've been cleaning things up.. not soon enough to make
life easy for anyone reviewing this (yet).

I have done a quick and dirty touchup to the patch however. As a result
you need to add fhandler_fifo.cc to Makefile.in. Lastly fifo.patch was
generated in the winsup dir.

Rob

fhandler_fifo.cc

fifo.patch

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/