McLemon, Sean wrote:
Having configured the Boost library for our embedded target, built Boost and tested a few examples successfully we'd like to incorporate Boost into our testing framework. I've a couple of queries, though:
1. When we run the run.py script it checks out a fresh version of Boost from SVN, perhaps clobbering local changes or at least wasting a lot of time before testing. The workaround suggested in the website (http://beta.boost.org/development/running_regression_tests.html) - that is passing "--user=" to the script - doesn't change this. Is there any way to prevent this happening reliably?
There's some option to keep the local source, something like --have-source, but I don't remember exactly.
2. Any executable we generate will need to be run using our command line simulator (i.e. we might create test.dxe, and run using jitccs532.exe test.dxe), how should I convey this to the test suite?
You should pass testing.launcher=jitccs532.exe to bjam command line.
3. Our simulator does support input redirection via the command line (i.e. $ yes | jitccs532 test.dxe) or command line arguments - does this mean that some of tests won't be usable?
"does", or "does not"? Most tests don't do any io or take any command line options. Some do, and surely won't work if your simulator does not provide that. - Volodya