Rendering Subversion version of HEAD

Anderson Lizardo lizardo at linuxfromscratch.org
Sun Jul 4 15:18:38 PDT 2004


On Saturday 03 July 2004 21:00, Kevin P. Fleming wrote:
> Anderson Lizardo wrote:
> > You mean moving that hard-coded "release table" to subversion properties?
> > If so, I think it's a good idea. I'm just finishing the new
> > render-lfs-book.sh now, It was simplified a lot, now it just needs to
> > receive a svn path (like "trunk" or "tags/v5_1_1") and an alias (like
> > "unstable" or stable"). By using properties, we can have the following
> > attached to e.g. the BOOK directory:
> >
> > stable=tags/v5_1_1
> > unstable=trunk
> > testing=branches/b6_0
>
> Yes, that exactly what I was thinking! However, in my mind it was going
> to be more "automatic", although that may not actually be possible... I
> was hoping the script could somehow scan the contents of the SVN repos,
> looking in trunk/ and tags/ for BOOK directories with a property of
> "render_to", or something like that. Whenever it finds a BOOK directory
> with that property, it would render the book into a directory in the
> lfs/view/... area named with the contents of the property.
>
> This would work nicely when the book moves to render-on-commit; when a
> commit occurs to the book, check the containing BOOK directory for a
> property called "render_to", and if it's there, re-render the book to
> the named place.

This seems to be possible. As render-lfs-book.sh already expects to receive 
the SVN path of the tree to be rendered, it can check that property before 
doing rendering. The only problem I can see is with the "repository scanning" 
you suggested. I suppose the script will become increasingly slower with new 
tags/braches added. As branching/tagging is a cheap operation on subversion, 
I foresee many of them coming ;).

IMHO that "scanning" would actually only be useful if we decide to continue 
using cron for running nightly builds. But with a post-commit process that 
would not be needed anymore, as is more resource-savy to only render the 
trees that have changed _and_ are marked with the "render_to" (or 
"render_me" :) property.
-- 
Anderson Lizardo
lizardo at linuxfromscratch.org
http://www.linuxfromscratch.org/



More information about the lfs-dev mailing list