Hi Rene, can you point me to an active resource for defining how test directories are meant to be implemented in boost? Thanks, Matt On 15/06/2016 3:14 a.m., Rene Rivera wrote:
As part of continued efforts to move testing to CI, to facilitate release management, and to provide more information to authors I've now implemented some testing changes:
1. Most libraries are now tested automatically if they have a usable test directory.
There are still libraries that do not follow the requirements of having all their tests in just the single test directory. You can see which ones those are here < https://github.com/boostorg/boost/blob/develop/status/Jamfile.v2#L101>. Library authors should IMMEDIATELY make changes to their test/Jamfile's to follow the library requirements for testing. Before the next release I plan to blindly remove that custom list of extra test targets.
2. Most structure library requirements are now checked both in the regular full testing and in the CI testing.
I've implemented a script to check for most of the library directory structure and integration requirements. The script runs as an additional CI job, currently in Travis. Here's one of those jobs < https://travis-ci.org/boostorg/boost/jobs/137506433>. At the moment it reports 17 libraries as having errors. There are also some warnings, i.e. not requirement violations but just some stuff that may be wrong. The script also runs as part of the general full testing and any problems are reported in each library's results as a single entry at the top of the tables. Here's how one of the test failures looks like < http://tinyurl.com/zbe9gal> [1] [2].
It would be great if library authors address those errors in some form ASAP. I know that for some of those reported libraries there will need to be some discussions as to how to best address the errors. Feel free to start additional threads for those as needed.
[1] < http://www.boost.org/development/tests/develop/developer/output/teeks99-09f-...
[2] Note there are currently some false failures in this aspect from a build system "bug". The bug is already fixed but tests will take some time to cycle to reflect the bug fix.