On Fri, 12 Oct 2012, Petr Viktorin wrote:
I purposedly went regexp way because of _("Additional instructions"). I
know that our testsuite is not passing when running it localized so it
may be a moot point but avoiding dependency on a localized content in
the test was one specific reason for the approach.
ACKed off-list, the test issue can be sorted out separately.
I think this should wait for a larger-scale, more thought-out rework
of the test suite.
At first glance, your approach seemed too ad-hoc and complicated. The
problem with complicated code in tests is that if code is
complicated, you need to test it -- leading to a loop.
(At second glance, the regexp actually turned out incorrect, but
that's less of a point.)
I'd be glad if we can come up with a good solution to testing i18n
issues everywhere, but this level of complexity in individual tests
is not it.
Sure, let's think about better way to make testsuite resistant to
environmental changes. Below, for the record, some of our ideas from
off-list discussion:
- force 'make-test' running with pre-set LC_ALL to something we expect
working;
- use special locale 'test.utf8' to force using 'test' translation.
test.po file would contain fixed translations that we can rely on in
the test suite;
- this special test translation can also test cases we know about as
being broken for lower/upper Unicode transformation (single-direction
changes like dotless i to I to i and others) where regressions might
appear.
--
/ Alexander Bokovoy
_______________________________________________
Freeipa-devel mailing list
Freeipa-devel@redhat.com
https://www.redhat.com/mailman/listinfo/freeipa-devel
