|
|
Boost Testing : |
From: Stefan Slapeta (stefan_at_[hidden])
Date: 2005-05-10 03:50:40
Paul Baxter wrote:
>
>
> Accompanying my last set of regression test results I posted details of
> my setup to this list.
> I'm sorry I didn't know about comment.html (documented somewhere?) which
> is why I posted some details to the thread.
> (http://article.gmane.org/gmane.comp.lib.boost.testing/829
> for who I am, and
> http://article.gmane.org/gmane.comp.lib.boost.testing/870 for the latest
> setup.)
you are right - it's not documented, we should fix that. just place a
file "comment.html" with some description in your regression directory.
>
> At the advice of Jeff Garland I updated to latest icc and changed from
> doing a default debug build [with v8.1.025 on early morning 3 May] to a
> release build [8.1.027 later that same day]. All running on WinXP pro
> (32 bit), AMD 2000+ XP.
>
please could you use the toolset intel-win32-8_1 then?
> I also have a new redhat4 x86_64 setup (P4 3.6 GHz) which I plan to use
> when I figure out how to get it to download the tests from cvs without
> the dreaded 'cvs lock held by ....'. This can also occasionally run some
> Intel for windows tests if you identify another useful configuration.
>
> I believe Stefan runs a debug build hence my change to release.
>
yes I do. an implicit hint that I should document my command line ;-)
> A long time ago I recommended the author of regex should take a look at
> www.boost.org and the rest, to John's credit, is history.
>
> As I have explained before, I am only a user and occasional tester but
> would welcome the chance to help in running the tests. I don't want to
> confuse the regression testing with my ad hoc results so I'd be happy to
> stop posting if needed. I tend to run the full tests as I've been caught
> by stale results in the past and although I don't run these regularly,
> if someone wants a test repeated I'll usually be able to do it within a
> day or so. (Private email request is OK)
>
> I made the suggestion before that if you want the variety of testing on
> many setups, perhaps you could provide a user set of results separate
> from the core largely automatic set of results. I certainly don't want
> to confuse other users.
>
Any help is welcome! (and it's never bad to have more results.) It's
just not completely clear if we should test release builds at all at the
moment, as there are many tests that generally fail for release builds.)
This is one of many points to discuss/resolve in our testing procedure...
Cheers,
Stefan