[Gnu-arch-users] Re: [BUG] "tla commit" vs. changes

[Julian T. J. Midgley]

In article <address@hidden>,
Johannes Berg  <address@hidden> wrote:
>-=-=-=-=-=-
>-=-=-=-=-=-
>
>On Mon, 2004-05-24 at 07:50, Jani Monoses wrote:
>> How is the out-of-date-ok case different for commit? People should know about
>> updating their project-tree before attempting commit, so if they don't it
>> means they know what they're doing.
>
>I think most people work against their own archive and are the only one
>to commit there and then only from a single source location, so it never
>happens that they're out of date.

I am certain that this falls into the class of argument that Tom
flamed me for earlier in this thread.

As it happens, I've instigated the use of arch at the company I work
for using a shared repository model (there is a single development
platform shared between multiple developers working on different parts
of the code, so having a single, shared development branch eliminates
needless merging and speeds development).

The consequences of committing modifications made to an out of date
tree are potentially severe (other people's code gets backed out or
broken without anyone becoming aware of the problem until some time
after the commit).  An empty (or incorrect) patch log is less of a
problem (although it would be nice if there were some way to
retrospectively modify a patch log when errors have been made).

All the best,

Julian
-- 
Julian T. J. Midgley                       http://www.xenoclast.org/
Cambridge, England.
PGP: BCC7863F FP: 52D9 1750 5721 7E58 C9E1  A7D5 3027 2F2E BCC7 863F