Scott Kostyshak wrote:

> Do you mean with the patch in this email? Or do you mean on current
> master, after commit 86325e50?

with the patch, as you found out.

> Yes we will have to discuss this. One issue is that you need a lot of
> dependencies to run all the tests. Another issue is that a couple of
> commits have broken many (100s) of tests:
> 
> edd37de8
> 8bd66109
> 
> Both commits did not introduce bugs themselves (although they might have
> introduced regressions from the perspective of the user) but they did
> expose current LyX bugs. I'm not sure what can be done here. Either we
> turn the tests off in order to get cleaner test output or we fix the
> exposed bugs:
> 
> http://www.lyx.org/trac/ticket/9633

I am a bit confused. I thought my fix which I just submitted would address 
this?

> http://www.lyx.org/trac/ticket/9744

I think the automatic part of this should go into 2.2.0. This is easy to do 
and safe and would both help us in testing and the users. I cannot comment 
on the parallel configuration of both font engines, since I do not 
understand the LaTeX font machinery well enough, I think here we should rely 
on Günters judgement since he knows this stuff well.

> Yes, Kornel has improved things a lot for this. To run with 8 cores you
> can do
> ctest -j8

Thanks, I'll try that.

> There are still problems with running the tests in parallel. I run them
> on just one core to be safe, but I think a reasonable approach is to run
> them on multiple cores, then re-run the failed test. So something like:
> 
> ctest -j8
> ctest --rerun-failed
> 
> In other words, I think that there is more of a concern of tests failing
> when they should pass than the other way around, when doing the tests in
> parallel.

OK, one could still use the parallel version for quick testing what a 
particular change does, and the serial one for final verification. This 
would already save time.

LyX is currently not very well suited for doing test driven development, but 
I can tell that (if the infrastructure is there) one gets good results very 
fast, and it even makes fun (despite being a buzz-word). Therefore I am 
trying to push for it.


Georg

Reply via email to