On 29 May 2014 at 8:29, Gennadiy Rozental wrote:
is one that several library owners have told me that they use instead of Boost.Test because B.T. kept breaking their builds and the documentation was not usable. Boost.Test does not change for many many years now
This is untrue. Here is an example of a recent Boost.Test change induced break: Note the #ifdef testing Boost.Test version: https://github.com/BoostGSoC13/boost.afio/blob/master/libs/afio/test/t est_functions.hpp#L160 Note the fact we need this at all is because Boost.Test won't timeout tests in a sane way (or at all on Windows), so we reimplement the functionality such that it actually works. I never got round to reporting this on the bug tracker, apologies.
* No one ever expressed a problem with Boost.Test being slow
Quite a bunch of people expressed frustration with the compile time at the C++ Now! 2014 workshop.
Do you have any numbers? How many seconds does it take to build Boost.Test shared library based test module vs. one built using other alternatives?
Boost.Test should default to not being header only. It's the only library in Boost I feel that way about. There is no good reason for a full fat unit test library to be header only. AFIO forces library only Boost.Test, and we saw a very dramatic improvement in CI turnaround times. A huge win. Niall -- ned Productions Limited Consulting http://www.nedproductions.biz/ http://ie.linkedin.com/in/nialldouglas/