On 12 Mar 2014 at 10:26, Borislav Stanimirov wrote:
At first that got me really excited, but as I found out after a short research, even Boost.AFIO's .travis.yml indicates that getting a Boost library to build with Travis CI is very hard, if at all reasonable.
You surprise me. Simply clone AFIO's travis.yml and adjust as fit, it's only Unix shell scripting. Note most of the tests are commented out because I have Jenkins CI doing all the testing.
Ah, my mistake then. Looking at the yml, with most of its lines commented out, it just seemed to me that you had spent some trying to get it to work and eventually gave up and switched to your own CI on your own server/machine (https://ci.nedprod.com).
I'll go add some explanatory comments, thanks for the tip. No, I am probably happier with the Travis CI output than the Jenkins CI output. Travis understands branches much better for example. Travis actually tells me useful information, whereas Jenkins tells me lots of stuff I don't need to know and forces me to click down multiple pages to get at what I actually want to know. If it weren't for needing Windows and BSD CI, I'd have stuck with Travis.
I'm going to try using it this weekend. Thanks
If you need any help, come back here. Myself and Antony Polukhin are very keen to get a lot more Travis testing into Boost. Best start is with new libraries. BTW, that CI caught my very first C++11 induced bug a few months back! The clue was that only very new compilers segfaulted i.e. if the compiler implemented rvalue refs v3, you got the segfault, otherwise not. Without the CI I'd have never noticed, nor had such a great clue as to what had gone wrong. Niall -- Currently unemployed and looking for work in Ireland. Work Portfolio: http://careers.stackoverflow.com/nialldouglas/