On Mon, Feb 21, 2011 at 12:19:34PM +0000, Philip Martin wrote:
> Stefan Sperling <stsp_at_elego.de> writes:
>
> > Fair enough, but why not write a specific test for this instead of testing
> > it as a side effect of making the code of other tests harder to read?
>
> The assumption is that translation only affects output to the user, not
> the internal behaviour of the commands. We have some translation tests
> based on that. To get full coverage of all the internal behaviour
> really requires that the tests can be run in a different locale.
I'd say it's an experiment with questionable gain for the price of
a lot of churn in existing tests. It might be difficult to come up
with an automated approach to substituting error messages with the
corresponding numbers. If this is done manually it will likely become
a time sink for the person who does it.
If, in spite of this, running the entire test suite in a different locale
is attempted, I'd prefer if the attempt was made on a branch instead of
trunk so that we can independently compare both approaches and come
to a decision based on these observations.
Received on 2011-02-21 13:31:40 CET