|
Boost : |
From: bill_kempf (williamkempf_at_[hidden])
Date: 2002-01-11 15:38:39
--- In boost_at_y..., Brad King <brad.king_at_k...> wrote:
> David,
>
> > I bet it's doable. I did a bunch of work on getting Jam to
generate
> > html for test result output already, but it's not integrated into
the
> > system yet.
> Perhaps. You'll have to run Dart and look at one of the generated
> files. I don't think the format is too complicated.
>
> > I'm not happy with any direction that requires specifying tests
in two
> > different languages. Now boost developers either need to install
Dart
> > or we have duplicated test specifications. I think either one is
> > unacceptable.
> We won't need duplicated test specs. The solution I was suggesting
was to
> have the Jam rules generate the DartTestfiles themselves. I'm
afraid that
> installing Dart is unavoidable unless you can duplicate the entire
> client-side with Jam code. Dart is fully interpreted,
so "installing" it
> simply means doing a CVS checkout or extracting a tarball.
However, tcl
> may need to be installed. I don't think this is too much to ask
from
> someone capable of coding for boost, and user's won't need to run
the
> tests.
Not too much in term of being able to work through how to get this
working. However, it can be considered too much to ask for the
installation of such a system. That's why Python was discounted in
favor of Jam. It's best if we have a single, light weight, non-
intrusive system for all of this stuff.
(I say that despite the fact that I have both Python and TCL
installed on my own systems. I'd actually love a Python based
solution for everything here, I just know that there are environments
where this will not be acceptable, and we don't want to disqualify
Boost in those environsments.)
> Another idea is to have some jam rules to actually execute Dart.
This
> could be used to get around the LD_LIBRARY_PATH problem. Jam could
have
> generated the list of tests into the DartTestfile, and then setup
the
> environment correctly before executing Dart. This way a test build
could
> be done like this (assuming BOOST_ROOT, ALL_LOCATE_TARGET, etc are
set in
> the environment already):
If that meant that writing tests was done once and TCL was not
required to run said tests this would be acceptable. But that's not
the case. As a developer I have to test my code, but if I'm in a
restricted environment where TCL is not an acceptable tool for
installation...
Bill Kempf
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk