On Thu, 11 Mar 2004, Norman Gray wrote:
> I've also realised (doh!) that because the generated configure file
> has baked into it a reference to the value you declare in
> STARCONF_DEFAULT_PREFIX -- this is part of the point of that variable
> -- the ./configure script is _always_ going to be specific to the
> developer's machine.
>
> When we get around to the business of rolling distribution tarballs,
> this will have to be taken into account (and documented), in one of
> a variety of ways: either setting STARCONF_DEFAULT_PREFIX to be /star
> (or perhaps /stardev if it's a beta of some sort), or ensuring that that
> variable is undefined and using a starconf which has been configured with
> './configure STARCONF_DEFAULT_PREFIX=/star'. (Et cetera: I'm writing
> this down partly as a memo rather than because anyone cares right now).
don't want to interrupt your train of thought but - does this mean
that users will be obliged to install in a given place? I'd hope
we could avoid that.
> But both this, and the fact that things like XXX_ERR and fac_xxx_err
> are to be generated, does further strengthen the case for keeping ALL
> generated files out of CVS. Keep adding them at present, but I think
> I will soon pick a time to go round all the components and remove
> Makefile.in and configure, and get starconf-validate to reverse its
> check-in test for those. I'll announce beforehand.
>
> Does anyone have any comments on this? The pros and cons are discussed on
> http://www.gnu.org/software/automake/manual/html_node/CVS.html -- as I've
> mentioned in passing before, I believe that the disadvantages of keeping
> generated files out, though persuasive in general, don't apply to us once
> we've decided to keep autotools in the repository and mandate their use.
I cast my vote in favour of keeping all generated files out of CVS,
if it can be made to work without too much aggro.
This is mainly on (strongly held, though not necessarily rationally
based) aesthetic grounds.
Mark
--
Mark Taylor Starlink Programmer Physics, Bristol University, UK
[log in to unmask] +44-117-928-8776 http://www.star.bris.ac.uk/~mbt/
|