Rather basic question about using mercurial in test/production set-up

[Marcin Kasperski]

> However I now have a situation where I do need to keep the testing
> code separate from production as it were.  It's my mail filter program
> which handles all my incoming mail.  Thus I can't just change it and
> have errors in it because I will lose mail.

Simplest approach is usually to use branches. For example:

- your „stable” code lies on branch named „production”

- you develop your changes on branch named „test”

- whenever you finish testing, you merge test to production
  to copy all those changes there

- when you have new ideas, you reopen test and go on.

Mercurial helps here a bit, as branches are „sticky” – you must
update to branch test on testing machine once, and future `hg update`
won't switch branch unless asked to. Similarly on prod,
once you `hg update production` once, future `hg update`-s will
stay on that branch (unless asked to)

Approach above can be made even simpler by using „default” as
development branch. Then you only need one named branch - `production`.
You hack freely, you update on testmachine to newest commit on default
to test, and from time to time you merge to production branch.