directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alex Karasulu" <akaras...@apache.org>
Subject Committing to Subversion (Re: Some questions about CI)
Date Tue, 18 Mar 2008 18:15:08 GMT
OK for starters I think you're wrong and I will show that.  But before doing
so, just for laughs, I want to ask you to set my sensitivity gauge on how I
shall proceed:

[ ] whisper sweet nothings into my ear, love me and be gentle
[ ] let me know what I'm doing wrong but don't embarrass me
[ ] be brutally honest potentially revealing my incompetence
[ ] rip into me like a porn star

:-D

Alex

On Tue, Mar 18, 2008 at 12:41 PM, Emmanuel Lecharny <elecharny@gmail.com>
wrote:

> Alex Karasulu wrote:
> >
> >
> >     >  - - How often shall a build be done (compile/test,
> sitegeneration)
> >
> >     We have many possible options. We tried something like kicking CI
> >     after each commit, but it leads to issues (usually, we don't commit
> >     code in one big shot,
> >
> >
> > Yes this does happen but it's bad practice on our part.
> I don't think so. First because when you commit, you usually have
> already checked locally that the server is ok (_usually_ => sometime,
> this is not the case ;). Second, because doing more than one commit
> allows you to comment more precisely what kind of modification you have
> done.
>
> I personnally don't really like to commit a big shot of code, unless it
> is really closely connected. But as I also fix some javadoc, bugs,
> warnings while browsing the code, I like to commit in smaller blocks.
> > I prefer a build on each commit so it's easier to catch the offending
> > commit and isolate it to a user who can be informed immediately while
> > they still have a mental stack in memory.
> If you kick a build after each commit, you may have many builds kicked
> when a lot of commits are done. I also think that it's quite rare that a
> commit break the build (it happens, say, every sic months ...), and when
> it does, being able to point the offending commit does not really helps
> to fix the breakage, because the offender is generally already sleeping :)
> >
> > I personally would like to know immediately when I goofed something
> > while that something is still in my head.
> Well, run the tests before committing should be enough, isn't it ?
>
> Don't get me wrong : I don't say that we should fragment commits as much
> as possible, nor I say that knowing which commit has broke a build is
> useless, I just say that a CI should be an airbag, when the integration
> is the safety belt. never commit without fasten your safety belt
> (-Dintegration test), and in case you crash the server, the the airbac
> (CI) may save your life !
>
>
> >
>
>
> --
> --
> cordialement, regards,
> Emmanuel L├ęcharny
> www.iktek.com
> directory.apache.org
>
>
>

Mime
View raw message