Re: [boost] ublas regression test problems

From: Beman Dawes (bdawes_at_[hidden])
Date: 2002-11-20 15:44:39

At 01:32 PM 11/20/2002, Toon Knapen wrote:

 >On Wednesday 20 November 2002 06:35, Aleksey Gurtovoy wrote:
 >> > P.S.: Why aren't there any regression tests for mpl?
 >>
 >> There are, they just not included into the batch. Probably they should
 >be;
 >> there are some issues, mainly with the number of tests (currently ~60 
and
 >> more to come) - simply putting those into results table would render it
 >> rather useless, IMO, and the corresponding increase in compile times
 >might
 >> also be a problem (for the tests runners). In any case, I almost never
 >> checking in anything before the regressions passed.
 >
 >Some have mentioned multiple times that the regression-test batch should 
not
 >
 >take too long to compile. I don't share this opinion. The more tests, the 

 >better and as long as the tests can be compiled in less then one night I
 >don't see a problem. For the status pages we (should) just have a cronjob 

 >that starts in the evening.
 >
 >As for reporting it's more difficult. Although Beman probably saw this 
one
 >coming and thus provided functionality to make a status page that only
 >reports the failed tests (as reported seperatly for windows 2K and vacpp)

I'm working on finishing the Filesystem library right now, as a 
prerequisite to upgrading the regression testing mechanisms, including 
figuring out how to handling every larger numbers of tests.

I'll ask for testing suggestions once the Filesystem work is done.

--Beman