Increasing reliability of main regression test

From: Beman Dawes (bdawes_at_[hidden])
Date: 2003-01-04 12:02:11

The following (or similar) scenario breaks the main regression tests fairly 
often. Twice this week, for example.

* Developer has small Jamfile in boost-root/libs/foo/build and another in 
boost-root/libs/foo/test. There is also a section testing foo in 
boost-root/status/Jamfile, often a cut-and-paste from 
boost-root/libs/foo/test/Jamfile.

* Developer makes some change in boost-root/libs/foo/build/Jamfile (or less 
often in other code) that requires a change in 
boost-root/libs/foo/test/Jamfile. Makes that change and tests, but 
completely forgets to change boost-root/status/Jamfile. Main regression 
test then breaks.

For the cases I'm familiar with, a fix would be for 
boost-root/status/Jamfile to be able to include the 
boost-root/libs/foo/test/Jamfile. I think several people (Gennadiy, Doug 
G.) have experimented with this.

We need to settle on an official way to do this.

Currently, such schemes break the status table generation because 
compiler_status.cpp actually scans status/Jamfile looking for test-type and 
some other more minor information. If we could figure a way to echo that 
information into the output (either to std::cout or in one of the residue 
files), it would make include schemes and compiler_status.cpp less fragile.

Ideas?

--Beman