On Fri, 16 Nov 2001 10:11:04 +0200
Michael Wood <mwood@its.uct.ac.za> wrote:
> On Thu, Nov 15, 2001 at 07:25:37PM -0600, Ben Collins-Sussman wrote:
> > Greg Stein <gstein@lyra.org> writes:
> >
> > > The whole "srcdir" file concept is really bogus... We should
> > > be passing the source directory on the command line. Apps
> > > that don't want it should ignore it. I had coded this at
> > > some point, but then somebody ripped it out. It really
> > > should go back in.
> >
> > It was taken out because nearly every test was complaining:
> > "warning: ignoring bogus argument". It seems awfully silly to
> > me that *every* test we write, past and future, now have this
> > obligation to expect (or deliberately ignore) a srcdir on the
> > commandline. It seems like an extra, unnecessary maintenance
> > hassle to me. :-(
>
> Any objections to using an environment variable?
Oh please no. Env vars work great under Unix but dealing with them
under Windows is like pulling teeth. You could not set one env var
anyway since each subdir is going to need its own srcdir. That is why
you need a different srcdir file in each subdirectory in the build dir.
For example, assume you root srcdir was /tmp/src/foo and you were running
in /tmp/build/foo/one and /tmp/build/foo/tmp. You would need to set
the var to /tmp/src/foo/one and then run the first test. You would then need
to set the var to /tmp/src/foo/tmp and run the second test. It would just be
a big pain since you would also need to pass the fully qualified path of the
python script on the command line:
cd /tmp/build/foo/one
setenv SRCDIR /tmp/src/foo/one
python /tmp/src/foo/one/test.python
Yuck, right?
Mo
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:48 2006