Re: [boost] Darwin regression tests

From: Chris Little (cslittle_at_[hidden])
Date: 2003-03-11 13:47:59

on 3/11/03 11:45 AM, Beman Dawes at bdawes_at_[hidden] wrote:

> At 11:01 AM 3/11/2003, Markus Schöpflin wrote:
> 
>> The must be something wrong with the current darwin regression tests,
>> AFAICT. I just completed a full regression run of the latest RC
>> version and I get far lower failure rates that the current regression
>> results indicate (9% failures instead of 45%).
> 
> My guess is that there was a problem with the configuration. Since the
> command lines reported by both sets of tests are the same, it may have been
> fixed but hasn't been reflected in the results because it wasn't a fix in a
> dependent file, so bjam never recompiled those tests. Or something like
> that.
> 
> The solution would be to clear out the regression target directories (or
> maybe just rerun with the -a switch.) Experiment with, say, just array1 to
> verify that takes care of the problem.
> 
> A 9% failure rate is much more what would be expected for GCC based
> compilers.

I emailed Marshall about this a couple of weeks ago.  There is something
wrong with the cs-Darwin.html file.  If gcc toolset failed a test then that
test is marked failed for the darwin toolset too but the fail on the darwin
toolset isn't an html link.  Also of the gcc toolsets tests are marked as
Fail but they aren't links either.

I agree with you that Marshall needs clean the target directory and rerun
the regression.

Chris