Re: [boost] Re: Regression testing enhancements

From: Beman Dawes (bdawes_at_[hidden])
Date: 2003-11-03 20:24:44

At 08:03 AM 11/2/2003, Aleksey Gurtovoy wrote:

 >David Abrahams wrote:
 >> Misha Bergal <mbergal_at_[hidden]> writes:
 >> > 2. The python test fail if run as a part of status/jamfile - runs 
fine
 >> >    if ran from libs/python/test (see log below)
 >>
 >> Baffling.
 >
 >It's easy to reproduce - include Boost.Python into the status/jamfile, 
and
 >you'll get what we are getting. We would appreciate if somebody with
 >adequate amount of bjam knoweledge could look at this in more detail -
 >apparently we don't have enough qualification to understand what's
 >happenning there, and certainly not enough time to get ourselves to that
 >level.
 >
 >>
 >> > 3. The requirement of unique test names across all libraries is 
really
 >> >    hurting (see http://tinyurl.com/t88o).
 >>
 >> I believe this is a problem in the table generation script, not in the
 >> build system, correct?
 >
 >Not exactly. 'process_jam_log' is at fault, here. Beman's reply was that
 >it's a feature, but it just doesn't seem to be practical or reasonable to
 >maintain that requirement.

Originally Boost.Test placed all results in boost-root/status/bin/xxx, 
where xxx was a name based on the test name. If two tests had the same 
name, they would have ended up in the same directory, and since the 
filenames were the same, results would have been overwritten. Thus the 
"rename" feature to prevent name clashes.

Now that the sub-include feature is being used, test names really only have 
to be unique within the particular library. So we ought to be able to relax 
the naming requirements for libraries which use the sub-include feature.

I'll try to run some tests to see what has to be changed to relax the rule.

--Beman