I am concluding for a windows user the tests are not really of any advantage. In short I am having trouble even making the harness run on windows. NOTE: I am using the windows SDK for compilation rather than MINGW32 and such like which ports the GNU tools to windows, I found those more bother than they were worth.
Also I am having trouble even getting nose to install on python3 on windows. I don't like the setuptools/distribute stuff (that was why I took a time away from python as they seemed to be the dominating installer system) and the nose docs indicates I can download the package and just use the setup.py script, but that seems to try and download distribute and then distribute fails to install when using python3.
Using python2 I do get nose installed, however when I try and run the runHarness.py script it complains that certain tables cannot be found. As I remember this was an issue with earlier versions of the harness. I think the issue is that the tables the tests rely on are split between the main liblouis tables directory and the tables directory of the tests directory. Running make check may set this all up for GNU tools users but doesn't help the windows user who might be using Microsoft tools for building.
I certainly could look at the tables issue, I feel though it doesn't really need a fix in the harness but rather a better plan on where tables for tests should be found (IE. should tests only rely on tables in the tables subdirectory of the tests directory?).
As for the nose issues, I am not even going near that, setuptools and distribute are projects I want to steer well clear of.
Michael Whapples On 13/06/2012 22:26, Mesar Hameed wrote:
Hi Vic On Wed 13/06/12,17:10, Vic Beckley wrote:Hi all, I have seen a tremendous amount of traffic lately on the harness test files.Yes sorry about that, hope it wasnt to irritating. we kept changing bits and peaces until we finally seem to have settled :) for now at least.I don't understand their purposeThe purpose is to have real input texts in a given language, and the matching expected (correct) braille output. For example english grade 2, word "the" should contract to dots 2346 The purpose of the harness is to have a list of input texts, single words, anything to trigger a contraction rule, and its known (manually checked) braille. Then when the harness runs we can quickly identify what words dont translate correctly, and how liblouis braille output is different from the expected. This should then gives us extra knowledge to correct liblouis opcodes/rules. It isnt something that will be used by everyday liblouis consumers, but more targeted towards table debugger/improver people.and how to use them.Assuming you have python, you probably also need to install the nose modules, in theory when you run make check it should all work.Are they of any value to me running strictly, for now, under Windows?Strictly not needed, but if you are interested in making sure your tables are correct/complete then you may consider giving us a hand. for a relatively small example you can have a look at the en-gb-g2 harness file Note that file does both forward and backward translation, but for a first step just considering forward translations is probably simplest. Thanks, Mesart For a description of the software, to download it and links to project pages go to http://www.abilitiessoft.com
For a description of the software, to download it and links to project pages go to http://www.abilitiessoft.com
