[Top][All Lists]

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

Re: before checking in

From: Eli Zaretskii
Subject: Re: before checking in
Date: Thu, 25 Oct 2001 11:59:03 +0200 (IST)

On Wed, 24 Oct 2001, John Wiegley wrote:

> Please check that all your changes compile without warnings before
> checking them into CVS.

I agree in principle that every change should be tested (and not only
by building).  But please note that it's not practically possible for
the developers to make sure _every_ Emacs configuration compiles
cleanly, because there are enough ifdef's in the code to make the
number of different configurations almost infinite.  There are system
dependencies, code that works only with certain toolkits and not with
others, code that is compiled on X but not on MS-Windows (and vice
versa), etc.  It is impractical to require the developer who makes the
change to test all those builds before committing.

There's a mailing list (address@hidden) to which you can
subscribe.  You will get email notifications about all commits to the
Emacs CVS tree, and you can then look closely at the changes in files
that interest you, and cry bloody murder when you see some potential
breakage in some configuration.

I suggested some time ago that every set of changes be posted
somewhere (probably here) and discussed before it is committed.  That
would give people a chance to catch possible problems on specific
platforms before they get into CVS.  But that idea was not too
popular, to say the least, so the emacs-commit list was created as a

> There's no reason why the code in CVS shouldn't always be able to
> build correctly.

There are lots of reasons.  The development trunk is by definition
unstable; the code is made stable during pretests.  That's life,
sorry; people who want stable code should not be using the CVS

reply via email to

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