Boost logo

Boost :

From: David Abrahams (dave_at_[hidden])
Date: 2002-12-22 20:09:45


Alkis Evlogimenos <alkis_at_[hidden]> writes:

> On Sunday 22 December 2002 12:39 pm, Beman Dawes wrote:
>> At 02:30 PM 12/22/2002, David Abrahams wrote:
>> >Alkis Evlogimenos <alkis_at_[hidden]> writes:
>> >> On Sunday 22 December 2002 03:51 am, John Maddock wrote:
>> >>> The current Linux regression test results appear to be going a little
>> >>> wrong, any idea what's up?
>> >>
>> >> I followed the instructions for the merge of the test-sane branch in
>> >> trunk.
>> >
>> >There should be no special instructions for this, since Beman has
>> >already made the merge. Just update on the main trunk.
>> >
>> >> I recompiled bjam and the regression tools, and pass --dump-tests to
>> >> bjam when running the test suite (should I pass it when I build the
>> >> libraries too?).
>> >
>> >That's not neccessary --dump-tests just prints information about the
>> >source files used in all declared tests to stdout. It's also not
>> >neccessary to separately build the libraries, since they'll be built
>> >automatically as part of the testing procedure.
>>
>> That's all correct, AFAIK. Exactly what is failing? Did you first delete
>> all the bin directories?
>
> I am not sure what is failing... John Maddock said that there are some
> problems with todays regression tests. I still didn't figure out what.
>
> To answer your question: all files that are not in CVS are cleaned up before I
> start the regression tests.

For those running regression tests (and most Boost.Build users), I
really recommend setting ALL_LOCATE_TARGET so that the products of
your build all go into a single directory tree that's not tangled into
your source tree. It makes it much easier, among other things, to
blow away the old built files.

However, killing the old build products is no longer neccessary since
sane_testing was merged into the main trunk. The whole point of that
new sanity is that outdated tests will always be re-run. The only
tests that will not be re-run are ones which succeeded in their
previous run, and whose dependencies haven't changed.

-- 
                       David Abrahams
   dave_at_[hidden] * http://www.boost-consulting.com
Boost support, enhancements, training, and commercial distribution

Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk