Re: [Boost-build] [1.34] Move to v2

From: Reece Dunn (msclrhd_at_[hidden])
Date: 2006-03-07 04:09:23

Vladimir Prus wrote:
>On Monday 06 March 2006 20:37, Thomas Witt wrote:
> > Could anybody give me a short status update?
>
>Sure. Reece has being running V2 tests with msvc and cw recently. We're 
>down
>to 6 V2-specific failures for msvc, three of them are due to Reece missing
>compression libraries, one almost surely fixed in CVS, and 2 remaining 
>Python
>failures being investigated

I have completed a run of cw-9.4 tests (using 'warnings=on', so you should 
actually be able to see the errors now!) and have started the msvc tests 
running.

NOTE: Having figured out how to set up a monitored build run (using bjam 
-l600), I am adding msvc-8.0 to the tests. This is because some of the 
date_time tests were asserting . (This is a known issue and is noted by Jeff 
Garland.)

NOTE: In the V1 regression run, the iostream compression library tests are 
skipped (leaving the entry white) if the libraries don't exist. This could 
be done using something like:

rule comp-test ( compression : sources )
{
   if ! $(NO_$(compression))
   {
      return [ run $(compression)_test : $(sources) ] ;
   }
   return ;
}

(this hasn't been verified, though). That would make it consistent with V1 
and cut the errors in half!

>For safety, I'd prefer to try some approached with Python tests and then 
>wait
>until Reece does another run with msvc, and after that we can start asking
>regular regression runners to switch. Because of the current state of MSVC, 
>I
>expect no problems, and the whole processs should be finished in a week or
>two.

As I mentioned above, I have started another msvc run now. Let me know when 
the other changes are in so I can restart it.

Having had several run throughs with running the regression tests, I am now 
at the point where I can run the tests once every day, provided that the 
BBv2 tests aren't broken.

BTW: if there is a BBv2 problem (e.g. like the previous ptr_container 
problem), this doesn't get logged by the regression script. It would be 
useful if this could be captured and displayed on the regression results 
table to save me or others from posting the issue directly.

- Reece