Re: [Boost-testing] Outdated Win32 results

Subject: Re: [Boost-testing] Outdated Win32 results
From: John Maddock (boost.regex_at_[hidden])
Date: 2012-12-21 04:31:56

>I run all my tests with the --incremental flag, could this be causing what
> you're seeing?

It shouldn't do so, it would certainly be worrying if it did :-(

Do you still have libs/multiprecision/test/test_arithmetic.cpp in the repo?

> Another potential issue, I'm bjam seems to be crashing part way through
> each run. I've been meaning to do some digging into this, but haven't had
> the opportunity.  It crashes after msvc-11.0's pool test running valgrind
> (this test fails, but I think it is whatever is *next* that is causing the
> crash). Here's my whole bjam.log file if anyone is interested (8MB
> compressed): http://boost.teeks99.com/temp/bjam.log.zip

OK, so it gets to valgrind_config_check and then bombs out - that's the last 
config check that's run, after that bjam should be dumping all of the:

boost-test(RUN) "accumulators/count" : "libs/accumulators/test/count.cpp"
boost-test(RUN) "accumulators/covariance" : 
"libs/accumulators/test/covariance.cpp"
boost-test(RUN) "accumulators/droppable" : 
"libs/accumulators/test/droppable.cpp"
boost-test(RUN) "accumulators/error_of" : 
"libs/accumulators/test/error_of.cpp"

stuff, so that's what's crashing.

If bjam is crashing like that, how do you get any results?  Or is the test 
matrix just reporting the last successful run which didn't crash?

The only suggestion I have is to not run all the compilers in one big run, 
in case bjam is running out of memory and then crashing.

Thanks, John.