gnustep-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Policy on checking in rebuilt configure


From: Ivan Vučica
Subject: Re: Policy on checking in rebuilt configure
Date: Mon, 5 May 2014 16:42:46 +0100

On Mon, May 5, 2014 at 8:13 AM, Fred Kiefer <address@hidden> wrote:
Not sure whether we need a new policy here. There is the simple rule that when you change configure.ac you also need to regenerate configure and check it in.

So that IS the policy. Good :-)
 
As far as I remember there was only one case in recent years where somebody didn't follow the rule but checked in a manually changed configure file.

I've done that at least a couple of times to avoid large diffs, which is why I'm asking now. 
 
Not having configure checked in is an option but it requires one additional tool installed on every machine where GNUstep is build. And it breaks the expectation that "configure; make; make install" builds a usable system.

I think that expectation is definitely reasonable with tarballs. However, it's also more than useless with VCS -- it's actively harmful by making diffs harder to examine, which is a problem if we ever start doing code reviews.

Expecting people who build from version control to have autotools installed is probably reasonable. And it also incentivizes us to do releases more often for the sake of people who don't want autotools installed in any shape of form.


On Mon, May 5, 2014 at 2:59 PM, Richard Frith-Macdonald <address@hidden> wrote:
It would be nice if we standardised on using the most recently released autoconf,  in order to keep diffs of the configure files to a minimum (I understand your distaste for large diffs).

Agreed. That however makes it painful -- how do we define 'most recently released autoconf'? I'd prefer not to have yet another tool installed *from source*.
 
BUT ... nobody really needs to review changes to configure, since it's trivial to regenerate it if there's a faulty/corrupt commit; so I don't think it's anything to worry about if we have occasional large diffs in the configure scripts.

Sure -- it's not a big problem. It is making a bit of noise if someone happens to be looking at the patches.


Note, I don't feel strongly about keeping 'configure' in VCS, just expressing why I feel it's unnecessary. If the agreed approach is 'just overwrite it', let's do that.

--
Ivan Vučica
address@hidden

reply via email to

[Prev in Thread] Current Thread [Next in Thread]