Boost Testing :
Subject: Re: [Boost-testing] Updated library_status for easier use..
From: Rene Rivera (grafikrobot_at_[hidden])
Date: 2015-04-30 11:46:30
On Wed, Apr 29, 2015 at 11:12 PM, Robert Ramey <ramey_at_[hidden]> wrote:
> Rene Rivera-2 wrote
> > Following up on the reorganization of the regression tools.. I'm happy to
> > say that part of the changes I pushed last night include a cleaned up and
> > easier to use library_status tool. Key changes, from the older version
> > the one Robert forked
> lol - I'm not sure that's the right term as I wrote it in the first place
Good point :-)
> > , are:
> > * Processing the jam log (i.e. process_jam_log) is integrated into it.
> I'm not at all sure this is great idea.
> What happens when there is not log to process?
Nothing is processed, as might be expected.
> It's not uncommon for
> me to run b2 - then find that there is a snafu of some sort. e.g. recently
> I ran with toolset=gcc-4.9 on my mac os system when I should have
> ran with darwin-4.8 Of course this created a column in the chart with
> 1000 lines of failures. To address cases like these it would be custom to
> remove all the directories for the bogus tests and re-run library_status
> which would re-build the chart.
> I don't see how one is going to do that now.
That would still work.. As processing the log just places XML files in
those same test directories you would delete.
> * Running b2 to build/run tests is also integrated into it (optionally).
> which sounds like another bad idea. I'm not sure how one passes all the
> command line parameters through library_status (I think library_status
> has some of it's own also) but at the least it requires an update in
> the documentation - which library_status hardly needed at all.
>From the documentation:
*regression-root*/stage/bin/library_status --b2 b2 --boost-root=../../..
But of course docs where minimal before, as they still are. So documenting
all the options in the HTML would be good.
> Which means that there's no external support script(s) needed and that
> > it's
> > fully cross-platform single executable. Documentation is updated at <
> > https://rawgit.com/boostorg/regression/develop/index.html> on how to
> > it and on using it at <
> > Obviously improvements from others welcome (as PRs).
> OK, good luck with this. Shouldn't be a big issue since as far as I know
> only two people
> besides myself (and now you) have shown any interest.
All I was trying to do was to make it easier by only needing *one* program
to worry about not a dance of three plus multiple per-platform scripts. So
perhaps you are right and I should just abandon this. I'll delete
library_status from the regression repo and you can deal with it on your
own. That way I can concentrate on replacing the regression reporting with
something modern and responsive.
But this does illustrate a big problem with boost tools. Here we've started
> out a simple idea
> and tarted it up to so that's not simple anymore. That's how our testing
> system got to the
> point where no who isn't required to use it does so. It's almost useless
> for most of us.
I could say a lot about that but I wont ;-)
> BTW - I still can't see the complete error message in the Dev matrix for
> library - this is holding me up in getting my tests to pass so I can merge
> to master.
> When is this going to get fixed.
I wasn't aware it was a problem that was holding anything up. Can you add
an issue here <https://github.com/boostorg/regression/issues> with a link
to the results that are problematic?
-- -- Rene Rivera -- Grafik - Don't Assume Anything -- Robot Dreams - http://robot-dreams.net -- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail