On Wed, Mar 14, 2012 at 9:00 PM, Les Mikesell <lesmikes...@gmail.com> wrote:
> On Wed, Mar 14, 2012 at 7:29 PM, Nico Kadel-Garcia <nka...@gmail.com> > wrote: > > On Wed, Mar 14, 2012 at 9:33 AM, Qazwart <qazw...@gmail.com> wrote: > >> > >> What if you deliver your source code *outside* of your revision control > >> system? This is quite common for open source projects. If I can't trust > the > >> build process to do a clean, I'm in trouble. > >> > > What if you don't use a build process? Just today, I deployed an > > "/etc/yum.repos.d" directory under Subversion, using an "svn checkout > > --force" command to put it under Subversion. This is certainly not under > a > > "build" process, and I would have *loved* to ahve an "svn clean" > operation > > So that's what, 500 bytes total if you had just deleted the whole > thing and checked it out clean? > > In this case, more like 4 K. But it also creates churn and would *break* yum during that few moments of deletion and re-install, and it would alter permissions, which is *another* set of issues. I've actually had this happen in worse fashion in Nagios and httpd, were configuration files are read out of a directory and any file called *.cfg for Nagios and *.conf for HTTPD ae read. Debris files are a problem, and easy for someone to leave behind. And re-deploying /etc/nagios is file ownership sensitive: typically, the HTTP password file is kept in /etc/nagios/passwd, and *THAT* needs to be legible to the "apache" user, although the /etc/nagios directory needs to be legivle to the "nagios" user. Re-deploy also requires network access. Whatever for? The local working copy knows its expected state!
