Re: More locking, sort of

[Greg A. Woods]

[ On Thursday, September 5, 2002 at 13:10:51 (-0400), address@hidden wrote: ]
> Subject: More locking, sort of
>
> Hi wizards.  I have what I believe to be a different variant of the
> perpetual question about locking.  I have a situation where, due to
> persistent problems with some hackers modifying some parts of a code
> base without understanding it first, we end up with broken code much
> more frequently than is justified.  A potential solution being bandied
> about is to lock out write access to certain subsets of the CVS tree,
> unlocking only when a change has been thought through carefully.

The better solution is to use a source code control tool which forces a
two-phase commit and ideally one which can enforce the result to be a
working code baseline that passes all existing regression tests.

The two-phase commit allows for senior reviewers to reject any change
before the code gets into the baseline.

Automated regression testing of the baseline on every commit ensures
other developers always have correct code to work with.

One such tool with these features is Aegis.

        http://www.canb.auug.org.au/~millerp/aegis.html

They can be hacked onto CVS, but not very well.

-- 
                                                                Greg A. Woods

+1 416 218-0098;            <address@hidden>;           <address@hidden>
Planix, Inc. <address@hidden>; VE3TCP; Secrets of the Weird <address@hidden>