Gennadiy Rozental wrote:
Robert Ramey <ramey <at> rrsd.com> writes:
The new system would only re-run the testing of library A when changes to library B have been merged to release.
Do you mean we should test against release brunch?
yes
I thought you previously refer to the last release.
If I did - that was a mistake. I mean we should test against the "next release". But..... Actually, from my perspective I view boost as a collection of 'releases" - one for each library. So for a given library - say boost/config.h which the the serialization library depends upon, In my brain the version on the release branch is the "latest release" In fact, when I make projects which use boost - generally small one-off projects - I just use my local boost release branch. So what boost calls the "next release" really is the "latest release" for me. Is everyone following that? So from time to time I update my local copy of the release branch and re-run my tests. If I had nothing else to do, I should probably just re-run all the "out of date" tests on my local environment. Conceptually, it's all relative. What I really need is to be able to build it locally from the release branch. It should be doable (and easy to do), and I think I remember that I have done it, it's just that, like all good programmers, I'm just lazy so for documentation I just use the latest one I can find. Robert Ramey