httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Pier Fumagalli <p...@betaversion.org>
Subject Re: Proposal: Future release strategies
Date Fri, 07 Sep 2001 01:39:14 GMT
"Ryan Bloom" <rbb@covalent.net> wrote:

> On Thursday 06 September 2001 11:47, Bill Stoddard wrote:
>>> Ryan Bloom wrote:
>>>> We have no control over APR.  APR will not make a release just because
>>>> the web server wants it to.  Apache needs to either use an already
>>>> released APR, or it needs to specify a date/time to check out APR.
>>> 
>>> We have no control over libc either, and yet we use that. Surely we can
>>> just consider APR a necessary external library that we link against -
>>> which has it's own release schedule - or is APR still too closely tied
>>> up with Apache for this to be practical?
>> 
>> I think APR is still changing too much for this to be practical. IMO, it
>> would be a serious mistake to attempt to ship apache httpd 2.0 w/o
>> including a copy of APR that has been tested with the server. APR is not
>> (yet) libc.
> 
> Thie big problem with this, is being able to reproduce from CVS.  It may mean
> that we have to include a timestamp in buildconf when we tell people to check
> out APR.  The problem is that if I go check-out httpd-2.0.30, I don't want to
> find a tag in APR that says   APACHE_2_0_30 and check that out, I want to
> grab APR_1_0_1.

Being one of the few who uses APR outside the scope of HTTPD (I know only of
Subversion ATM using the lib for things different from the web server), yes,
I agree that it would be so cool to have an APR tag to rely on to reproduce
past builds. But I see Bill's point on saying that APR is not ready for
prime time yet, putting a tag right now would be (IMVVVVVVVHO) as checking
out with a specified date.

What I do when rolling a "milestone" of WebApp is that I "declare" a
date/time stamp I am more or less sure is working, testing it out, make a
snapshot, and include it with my source distrib... It has worked pretty well
in the last few months, and I couldn't ask for more since I was already able
to remove any possible "#ifdef NAME_YOUR_PLATFORM" from my code...

And if there's a bug, oh well, it's still alpha code :) Check out HEAD and
redo the whole thing again (takes me less time than figuring out all
platform dependancies anyway!)

    Pier


Mime
View raw message