Re: [jamboost] Small patch for BoostBuild.py

From: Niklaus Giger (NiklausGiger_at_[hidden])
Date: 2005-03-11 13:49:10

Am Freitag, 11. März 2005 09.16 schrieb Vladimir Prus:
> On Thursday 10 March 2005 20:48, Niklaus Giger wrote:
> > I wondered why the point 2 in test_system.html describing
> > run_build_system didn't work. It says "Changes to subdir, if it is
> > specified. If it is not absolute path, it is relative to the working
> > dir."
> >
> > After peeking into the source code this small patches allowed me to test
> > scripts directly in their directory without having to create a temporary
> > copy.
>
> I see, you want to use some existing directory for testing, right?
Yes, because there are small scripts but big directories.
> > Otherwise the test_system is very good for my needs and my tests begin to
> > consume hours of CPU time,
>
> Do you mean you're writing some tests of your own using the BoostBuild.py
> module?
Yes, I just used the framework for my own infrastructure. I use mostly the 
Perforce Jam. Once I have working regression tests we might switch more 
easily to Boost.jam.

> Two points:
>
> 1. Please send patches as unified diff. ("diff -u" or "cvs diff -u").
> 2. I'm worried about the different behaviour of relative and absolute paths
>
> - if you specify relative path, we track the changes to the entire
> working directory
> - if you specify absolute path (even if it points inside the working
> direcotory), we track changes only to the specified directory
>
> Maybe, a better approach would be:
>
> - outlaw absolute paths as value of 'subdir'
> - allows to specify working dir, but adding 'workdir' argument to
> __init__ and passing it to TestCmd.__init__
>
> Or, in fact, you can do this:
>
> t = Tester()
> os.chdir(t.original_workdir)
>
> just like project_test1.py does, but it's ugly.
>
The second approach is an ugly hack. I didn't manage to finish a working 
solution using your first approach, which is a lot cleaner. I will submit you 
in the next days a patch (in unified diff format) with the changes in the 
code, a test for the new feature and the corresponding changes in the html 
documentation. Tested at least under cygwin and GNU/Linux on PowerPC.

Best regards

-- 
Niklaus Giger
Wieshoschet 6
CH-8753 Mollis