Re: [PATCH] New margin handling - final version (updated)

[Neil Puttock]

2009/9/11 Michael Käppler <address@hidden>:

> ...however, I don't really understand the benefit of splitting the regtest.
> Do you propose to have separate files for each combination of settings?

It's recommended at the top of the regression testing page.

In your snippet, there's a lot of markup documenting the desired
behaviour, which would probably be better off moved into texidocs.

> And if warning-as-error was set to true, the regtest would fail. I don't
> think that is intended, since the >absence< of the warning would be wrong
> behaviour, not the appearance.

Patrick explains this more clearly.

Since we will eventually want 'warning-as-error to be set, any test
which should produce a warning message must have this option unset
(which is a good reason for hiving it off to its own test, in case
anything else breaks in the future).

Regards,
Neil