[Boost-testing] Improving turnaround time for regression tests? A simple idea...

From: Doug Gregor (dgregor_at_[hidden])
Date: 2005-04-22 08:56:55

I was looking at the list of regression results submitters and noticed 
that we have a few cases where the same compiler/platform is being 
submitted twice. That's usually a good thing... it makes sure minor 
variations are tested, providing better coverage. However, when one of 
the submitters is running tests for lots of compilers it might be 
better to only have the more-lightly-loaded tester run tests for that 
compiler. For instance, both RudbekAssociates and MetaComm are running 
tests for vc-7_1 on Windows, but MetaComm's tests lag farther behind 
because they're running tests for 8 (!) compiler/stdlib combinations. 
Similarly, RSI-SoL and Martin Wille are both running GCC 3.4.3 on 
Linux, with the latter running more tests and therefore lagging further 
behind.

I think coordination to distribute testing as much as possible would 
help push the release forward, because we can give developers better 
turnaround times on regression test results.

        Thanks!
        Doug