Re: [Boost-testing] Regressions with buildbot?

From: Aleksey Gurtovoy (agurtovoy_at_[hidden])
Date: 2007-01-17 22:53:38

Roland Schwarz writes:
> As I have already expressed in another post I have the feeling the
> regressions currently don't give a perfect picture about sanity of
> the boost lib. I might be wrong however.

You are right, they don't. See
http://article.gmane.org/gmane.comp.lib.boost.testing/3854.

> I doubt that the regression.py really is very failsafe, 

Meaning?

> and does not need much operator intervention.

There are known issues in this area and known solutions
(http://thread.gmane.org/gmane.comp.lib.boost.testing/2741/focus=3068
and
http://thread.gmane.org/gmane.comp.lib.boost.testing/3818/focus=3832). In
short, the above is not true.

> I never found out how the annoying message boxes, that hold the test
> are supposed to be handled. Can anyone explain please?

Please see the above.

> From a short browsing through the docs of buildbot I have seen, that at 
> least there is a timeout with promise to kill the session if nothing 
> received for timeout-period.

$ python regression.py 
Usage:
        regression.py [command] options

[...]

Options:
        --runner        runner ID (e.g. 'Metacomm')
        --tag           the tag for the results ('CVS-HEAD' by default)
        --local         the name of the boost tarball
        --comment       an HTML comment file to be inserted in the reports
                        ('comment.html' by default)
        --incremental   do incremental run (do not remove previous binaries)
        --force-update  do a CVS update (if applicable) instead of a clean
                        checkout, even when performing a full run
        --monitored     do a monitored run
        --timeout       specifies the timeout, in minutes, for a single test
                        run/compilation (enforced only in monitored runs, 5 by
                        default)
        ^^^^^^^^^

-- 
Aleksey Gurtovoy
MetaCommunications Engineering