[lsb-discuss] Where do I find the information on "proper format for integration with LSB"

Wichmann, Mats D mats.d.wichmann at intel.com
Tue Jul 31 18:02:45 UTC 2012


On Tue, Jul 31, 2012 at 11:35 AM, Petrie, Glen
<glen.petrie at eitc.epson.com> wrote:
>
> All
>
>
>
> In connection with SANE integration in LSB 5.0, there needs to be a test suite.   The SANE Library has a test suite for the library; however, the request in the LSB documentation (see below) is for a test suite that is proper formatted for integration into the LSB test suite.
>
>
>
> Since I have not integrated with LSB before, can someone point me to the "instructions" or a very good example of a proper formatted test suite


I think you'll get more detailed answers in a bit.  The two short answers for
the preferred approaches are:

(a) assertion-based tests
(b) results in the  TET format

For the former it means tests which make a claim based on the documented
API behavior, and then report a result of the form pass/fail/couldn't
run the test.
Such assertions are typically both positive and negative in nature - that is,
testing not only that things that should complete do so without error and with
correct results, but also that things which should cause errors do, and cause
the expected ones.

We have several different toolkits and frameworks, the common feature of nearly
all is the tet output format (formally described in
http://tetworks.opengroup.org/documents/3.8/pguide.pdf)
We have simple kits (e.g. in python) which can emit stuff in the right format
without deeper understanding of the internal test format, all the way
to actually
using the TETware API in the full way it was intended (but that implies having
TETware actually drive your tests, which you may or may not want depending
on what the existing setup is).  We also have some tests which use a richer
and more nuanced reporting format.

I say "preferred" because while the overall testing setup works best with
something which fully conforms, some tests that just do enough to be able
to have their results read is far better than zero tests which "do everything
in the standard way" :)

I think we've got some wiki pages on the topic.


More information about the lsb-discuss mailing list