Re: [Boost-testing] [boost] Regression tests broken

Subject: Re: [Boost-testing] [boost] Regression tests broken
From: Mateusz Loskot (mateusz_at_[hidden])
Date: 2019-02-19 11:58:43

On Tue, 19 Feb 2019 at 12:23, Mateusz Loskot <mateusz_at_[hidden]> wrote:
> On Tue, 19 Feb 2019 at 11:49, Hans Dembinski via Boost <boost_at_[hidden]> wrote:
> > > On 19. Feb 2019, at 04:29, Peter Dimov via Boost <boost_at_[hidden]> wrote:
> > >
> > > Tom Kent wrote:
> > >> All-
> > >> Sometime around 7pm yesterday (2-17) a commit went in (master and develop?, regression scripts?) that broke all my regression runners (linux and windows).
> > >
> > > It's this one: https://github.com/boostorg/histogram/commit/9abbe46e3dd57e2b55f7ab52eaed15994f76ae52#diff-42dd6ec1330a7c47aaccf2ab2b8f1e02R12
> >
> > so it was me, sorry. This error didn't show up in my unit test. It seems to come from my `coverage` variant, which I copied and modified from beast, where it seems to work fine.
>
> You seem to be running `b2 -j2 -q toolset=gcc-7 coverage`, while it
> should read `variant=coverage`
>
> AFAIS, Beast's `.travis.yml` sets `VARIANT` and other env variables, then calls
> `libs/beast/tools/retry.sh libs/beast/tools/build-and-test.sh`
> which composes the `b2` command line

I take that back. Boost.Build fu shortage here.
Your use seems fine [1], ass we've got it clarified on Slack.

[1] https://boostorg.github.io/build/manual/develop/index.html
"The release and debug that we have seen in b2 invocations are just
a shorthand way to specify values of the variant feature."

Best regards,
-- 
Mateusz Loskot, http://mateusz.loskot.net