Uwe Stöhr wrote:

> I just sent the complete patch. Could you please apply it and test. I am
> pretty sure it passes all tests but since i cannot perform them on my
> own...

I tested it, and the tests pass. This does unfortunately not say much, 
because we do not have any tex2lyx test that uses the modernCV layout, so 
even if tex2lyx would output broken modernCV documents, we would not know. 
However, since tex2lyx uses the layout file machinery of LyX, layout file 
updates are very unlikely to break tex2lyx if there is nothing hardcoded for 
the changed layout file. I am very sure that I did not write any hardcoded 
modernCV stuff for tex2lyx, so if you did not do that either then I think we 
can be sufficently sure that the changes are OK.

> p.s. A general thought: In my world the IT is extremely fast: one gets
> updates all the time.

Please remember that this looks quite different for some other people. We 
cannot assume that every user gets updates in this way.

> If I deny for example a Java update, they block my
> PC until this is in. I get every Monday the latest LaTeX-packages from
> MiKTeX, Windows 10 is now also adding features all the time. So LyX's
> model is a bit out of time.

I think LyX would be perfectly suitable for frequent updates in theory, and 
I'd like to see those as well one day. For example, all those small new 
features like the one we are discussing in this thread, would not need to 
wait for any big release. If you look how other projects work which use 
"Continous Delivery" (this is how the practice is called if you want to use 
a buzzword), then you see that the most important points are

1) automation
2) automation
3) did I mention automation yet?

If you do not have a very good test coverage (small and big tests, unit 
tests, functional tests, integration tests), and if you do not have an 
automated build server that continously builds the software for all 
platforms (in our case this would be (Windows, Linux and OS X) and runs all 
the test after each build, then you cannot do continous delivery. If you try 
it nevertheless you will piss off your users, because regressions will be 
introduced, and you cannot afford to do the manual testing we currently do 
for each release.

This is one reason why I try to convince everybody that tests are needed and 
your friend, not your enemy. If we do not increase our test coverage (and 
improve the test infrastructure), we will never be able to go towards 
continous delivery.


Georg


Reply via email to