Re: mod_fastcgi

Stanley Gambarin (stanleyg@cs.bu.edu)
Fri, 12 Sep 1997 16:35:34 -0400 (EDT)

Date: Fri, 12 Sep 1997 16:35:34 -0400 (EDT)
From: Stanley Gambarin <stanleyg@cs.bu.edu>
To: Gerardo Diaz Cuellar <gdiaz@blackbox.cegs.itesm.mx>
Subject: Re: mod_fastcgi
In-Reply-To: <Pine.NXT.3.91.970912131723.5526A-100000@blackbox>
Message-Id: <Pine.GSO.3.95q.970912163312.6814A-100000@csa>



=09Like I had mentioned before... there was only a
limited amount of testing of the C code.. I have not seen the
code in half year, so I do not remember what calls what..
I am gonna take a look at the code, but its gonna take some time...


=09=09=09=09=09=09Stanley.

On Fri, 12 Sep 1997, Gerardo Diaz Cuellar wrote:

> I thought that C calls directly the FCGI_ routines, but according to you=
=20
> that is not the case, am I correct?
>=20
> I so, then C calls FCGIX_ and should not have any problems. am I correct=
=20
> up to here?=20
>=20
> but FCGI_stdio.h has the FCGI_ routines that are the ones that the=20
> examples use (they included FCGI_stdio.h) , and if you tested C and=20
> theres no problem then what=B4 happening with C?
>=20
> regards
>=20
> -Gerardo=20
>=20
> On Thu, 11 Sep 1997, Stanley Gambarin wrote:
>=20
> >=20
> > =09It looks like developer's kit does not have a fix for accept()
> > problem.  In my somewhat limited testing back at OpenMarket, I was not
> > able to reproduce this problem with C, however, the problem was evident=
=20
> > with Perl.  What I think is happening is that FCGX_* routines are OK,
> > however, FCGI_* routines, which are built on top of FCGX_* ones do
> > have serialization problems, however, it looks like only in Perl. =20
> > Sfio/Fcgi module writes its own FCGI_* routines, using FCGX_* ones=20
> > (correct me if I am wrong), so if you use a Sfio/Fcgi perl module,
> > you should be ok.  C is also ok, but i guess it needs more testing...
> >=20
> >=20
> > =09=09=09=09Stanley.
> >=20
> > On Thu, 11 Sep 1997, Gerardo Diaz Cuellar wrote:
> >=20
> > > Ups.=20
> > >=20
> > > That means that the devkit is not fix? then programs directly to the=
=20
> > > devkit will still experience these problem (programs made in C). Do=
=20
> > > anyone knows when in OpenMarket going to fix this on the devkit?
> > >=20
> > > By the way if the perl module uses the devkit how come the perl modul=
e=20
> > > (sfio) is correct and the devkit is not? Exactly how does this calls =
take=20
> > > place?
> > >=20
> > > On Thu, 11 Sep 1997, Jonathan Roy wrote:
> > >=20
> > > >=20
> > > >   I had thought the FCGI perl moduble was built from the dev kit av=
ailable
> > > > on the web site, and the module does include the locking around the=
 initial
> > > > accept() and the fix to prevent lots of processes from incorrectly =
dying
> > > > off at startup time.
> > > >=20
> > > >   But actually, I looked back over my email archive and it appears =
the
> > > > patches were applied manually to .31 since they hadn't shown up in =
the
> > > > devkit at that time. If they still aren't in the devkit, I guess we=
'll just
> > > > have to fend for ourselves. :(
> > > >=20
> > > > -Jonathan
> > > >=20
> > > >=20
> > > > --
> > > > Jonathan Roy - roy@idle.com - Idle Communications, Inc.
> > > > Idle Communications, Inc. accepts contract programming
> > > > work for general purpose tools (or CGI) in Perl/C/C++.
> > > >=20
> > > >=20
> > >=20
> >=20
> >=20
>=20