httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From ras...@bellglobal.com
Subject Re: cvs commit: apachen STATUS
Date Thu, 08 Jan 1998 20:53:16 GMT
>     1. patch code
>     2. test patch until satisfied
>     3. cvs diff -u >neato.patch
>     4. review neato.patch, possibly edit out things you don't mean to
>     commit 5. cvs commit src/CHANGES src/foo/bar.c
>     6. goto 1 because you've still got way more time to play today
> 
> and for the minimal reviewer:
> 
>     1. read about patch on apache-cvs
>     2. cvs update, rebuild
>     3. test
>     4. goto 1 because you've still got more time to play today

This is how the PHP3 development cycle works.  There are 6 core developers
and about 10 other regular contributors.  Everyone has commit priviledges.  
We occasionally need to back out an ill-advised commit, but 99% of the time 
the commits are either fine or a little bit of tweaking by one of the core
guys makes it fine.  The only thing we have any semblance of voting on is
high-level concepts.  And it isn't really a vote, just some heated 
discussion back and forth until mutual agreement is reached.

And yes, the HEAD CVS doesn't always work very well.  Sometimes it doesn't
even compile.  But that also means that when there is a problem in a 
patch, it is discovered right away while the originator is still around and
has the code fresh in his mind.  I have found that with the Apache cycle, 
because it takes so long for a change to go through, it is difficult to
debug resulting problems.

I am not saying that this is necessarily the way to go for Apache, but I
do agree with Dean.  The current setup sucks.  I for one would be testing
a heck of a lot more patches if I could just cvs update and recompile.  I
don't have the cycles in my day to manually patch things.  That's 
obviously how we are finding most of the problems right now anyway.  Lots of
patches that get +1'ed end up causing things to break anyway and this is
caught months later instead of right away because of the current setup.

-Rasmus

Mime
View raw message