Re: [Boost-testing] Re: Intel 7.1-linux tarball results

From: Martin Wille (mw8329_at_[hidden])
Date: 2004-11-16 14:03:51

[not crossposted to the developer list]

Robert Ramey wrote:
> Isn't this a link from the 1.32 release package.  Can't the link just be
> left and point to the updated results?  In fact, isn't possible that a newer
> version of a compiler (e.g. gcc 3.4.3?) comes out and we want to re-run show
> updated test results eventhough the 1.32 package is frozen.  I don't see
> this has having anything to do with release drafts.

FWIW, I am already testing gcc 3.4.3 (the results will be available in
a couple of hours).

Robert raises an interesting point. I also thought it would be
nice to have an environment that would allow for test results to
be published incrementally. This would yield several advantages:

1. Partial results are available earlier

2. Tests can be run in parallel on similarly set up boxes without
    having to use different runner-ids.

3. Bogus results can be replaced without having to rebuild all
    the results for the entire set of toolsets.

4. Several smaller advantages, e.g.: being able to interleave
    uploads with running the tests, having more points at which
    testing can be interrupted without having to rerun anything.

Such an environment would consist of two result sets per runner
id. One set is considered "final" while the other is considered
"being built". Results would be uploaded per toolset into the
"being built" set. Eventually, that set would be committed
resulting in a new "final" set and a new, empty "being built"
set would be created.

I think this environment can be built without changing anything
in Boost.Build.

Regards,
m